Elucidating precursors of volcanic eruptions helps us understand the preparatory processes of the upcoming eruption. Previous studies reported various precursors of eruptions, including increased seismicity, amount of volcanic gas emission, and temperature in geothermal areas (e.g., Stix et al. 2018). Initiation of these precursors varies from years to minutes. For example, magma accumulation at shallow magma storage can precede an eruption by more than ten years, increasing seismicity often precedes an eruption by a few weeks, and tilt changes often precede an eruption by only minutes (e.g., Bell et al. 2021; Kato et al. 2015). A decrease in deformation rate or seismicity preceding volcanic eruptions has also been reported (Sigmundsson et al. 2022). Some eruptions lack any precursory signals because of their small amplitudes, the lack of detection ability, or the inexistence of the precursors (Lesage et al. 2018; Maeda et al. 2015).
Variation in the deformation rate is one of the indicators of an imminent eruption: an increase in the deformation rate implies an increase in the vertical or horizontal influx of volcanic materials (Kato et al. 2015; de Moor et al. 2016; Ardid et al. 2022; Smittarello et al. 2022). Geodetic instruments such as Global Navigation Satellite System (GNSS), tiltmeters, and Synthetic Aperture Radar (SAR) can detect the precursory overpressure of the volcanic system by observing its inflation. SAR is superior in detecting small-scale deformation signals or deformation in inaccessible areas, such as craters or geothermal areas. Several studies have succeeded in detecting local-scale ground deformations before eruptions using SAR images and suggested that the overpressure of the volcanic system commenced several months before eruptions (Hamling 2017; Kobayashi et al. 2018; Narita et al. 2020). The high spatial resolution data derived from SAR images helps understand the spatiotemporal feature of the ground deformation and the geometry of the deformation source.
Shinmoe-dake volcano is one of the active volcanoes in southwestern Japan and a part of the Kirishima volcanic complex (Fig. 1). Recent major eruptions of Shinmoe-dake volcano were in 2011, 2017, and 2018 (Nakada et al. 2013; Yamada et al. 2019). The 2011 eruption started with a subplinian eruption with the ejection of altered materials. A new vent was formed at the west side of the crater during the subplinian eruption, followed by the lava ejection from the crater’s center. SAR intensity images gave the amount of extruded lava at the summit 1.5×107 m3 (Kozono et al. 2013; Ozawa and Kozono 2013). The extruded lava became a dome shape within the crater. The lava dome inflated from the termination of the 2011 eruption until the middle of 2016 (Miyagi et al. 2014; National Research Institute for Earth Science and Disaster Resilience (NIED) 2017). Broad-scale inflations with 4 cm of line-of-sight (LOS) change and a spatial extent of 10 km located 5 km to the west-northwest from the crater preceding the 2011 eruption of Shinmoe-dake volcano (Miyagi et al. 2014). Following the eruption in 2011, the broad deformation signal transitioned to deflations at a similar spatial extent. This deflation implies a decrease in pressure caused by magma extrusion from the reservoir.
The 2017 eruption, which started on 11 October, was mainly accompanied by ash emissions (Japan Meteorological Agency (JMA) 2018). Although a new vent was formed during the 2017 eruption at the east side of the crater, the eruption did not change the shape of the lava dome (Additional File 1: Figure S1). GNSS baselines crossing the Kirishima volcanic complex (> 10 km) started to extend in mid-2017, suggesting an overpressure of the magma source (Yamada et al. 2019). The 2017 eruption made no significant changes in the GNSS baselines, while they extended with acceleration over time toward the following eruption in 2018. The seismic background noise increased a few months before the 2017 eruption and was kept high until the eruption. This observation suggests a long-term degassing from the magma within the conduit connecting the surface and the shallow magma reservoir (Ichihara et al. 2023a, b).
The 2018 eruption, which started on 1 March, was accompanied by both ash emission and lava extrusion. The time series of GNSS baselines crossing Shinmoe-dake volcano showed a baseline shortening during the early stage of the eruption when volcanic ash was mainly emitting. In contrast, the GNSS baselines turned to extension after the early stage of the eruption (Yamada et al. 2019). The lava was ejected from the vent from the east side of the crater after the ash emission phase and covered the pre-existing lava dome, which was formed during the 2011 eruption (Additional File 1: Figure S1; Ichihara et al. 2023a).
During the observation period of this study, Iwo-yama volcano, located 5 km to the northwest of Shinmoe-dake volcano, also experienced a steam blowout from the vent in April 2017 and a phreatic eruption in April 2018 (Narita et al. 2020; Tajima et al. 2020). These eruptions are related to the overpressure of the hydrothermal system beneath Shinmoe-dake and Iwo-yama volcanoes across the Kirishima volcanic complex (Narita et al. 2020; Ichihara et al. 2023a).
The spatiotemporal characteristics of ground deformations at the Kirishima volcanic complex (southwest Japan) from 2006 to 2019 have been reported from satellite SAR images (Yunjun et al. 2021). Notwithstanding previous studies, further discussions of the spatiotemporal features of the precursory ground deformation of the Shinmoe-dake volcano eruptions are required. Ground deformations at the crater especially contain crucial information about the pressure conditions within the volcanic conduit and the shallow part of the subsurface volcanic system. This paper aims to investigate the local-scale ground deformation at the crater before the Shinmoe-dake volcano eruptions in 2017 and 2018 based on the analysis of satellite SAR images. Through estimating the pressure source geometry and engaging in discussion, we aim to update the understanding of the preparatory processes of eruptions of Shinmoe-dake volcano.
SAR image processing and SAR time-series analysis
To detect the ground deformation related to the eruptions of Shinmoe-dake volcano in 2017 and 2018, we processed L-band SAR images acquired from Phased Array L-band SAR-2 (PALSAR-2) onboarding Advanced Land Observation Satellite-2 (ALOS-2) satellite between February 2016 and February 2018 and C-band Sentinel-1 image acquired between January 2017 and February 2018, respectively (Fig. 1; Additional File 1: Table S1). SAR images were processed using the GAMMA software (Wegmüller and Werner 1997). Topography-correlated phase changes were corrected using the Digital Elevation Model (DEM) with a spatial resolution of 10 m from the Geospatial Information Authority of Japan, which updated DEM around the Kirishima volcanic complex in 2016, before the eruption in 2017. Phase unwrapping was done with the minimum cost flow algorithm (Costantini 1998). Pixels with coherence below 0.1 were discarded as unreliable. We generated InSAR images from Sentinel-1 images with a temporal separation of less than 36 days to avoid decorrelation problems and those from PALSAR-2 images from all possible pairs.
We created two subsets of LOS change time series, one before the 2017 eruption and the other between the 2017 and 2018 eruptions, to avoid the contamination induced by decorrelation noise associated with the eruptions. We applied multi-temporal InSAR analysis to all Sentinel-1 and PALSAR-2 images acquired before the 2017 eruption (Berardino et al. 2002; Schmidt and Bürgmann 2003). The stacking approach was also applied to PALSAR-2 images acquired between the 2017 and 2018 eruptions because of limited available images. We applied the Laplacian operator for the multi-temporal InSAR analysis to temporally smooth LOS changes. The strength of the Laplacian operator was optimized using the L-curve criteria (Hansen, 1992; Additional File 1: Figure S2). Errors in DEM data were also simultaneously estimated with the spatiotemporal variation of LOS changes based on the perpendicular baselines of SAR images (Fattahi and Amelung, 2013). For the LOS change preceding the 2017 eruption, we applied multi-temporal InSAR analysis to PALSAR-2 images acquired since the beginning of 2016 to improve the measurement accuracy for SAR time series analysis by increasing the number of images. We then isolated the time series of PALSAR-2 LOS changes between the last data acquired in 2016 (Path 23: November 14, 2016; Path 130: December 15, 2016; Path 131: December 6, 2016) and the last data before the onset of the eruption in 2017 (October 11) from the entire LOS change time series.