Rainfall-induced Gravity Movement of the Unzen-Fugendake Volcanic Dome Analysis combining Ground-Radar Interferometry and XRAIN Rainfall radar system

Lava domes created by volcanoes often cause pyroclastic and debris flows, which have a significant impact on the surrounding infrastructure and population and have been the subject of much research. However, because volcanic domes tend not to survive the eruptions that form them, the instability of domes that survive eruptions such as Unzen Fugendake in Japan is both a poorly understood process and a danger. Therefore, the present contribution aims to (1) Displacement and precipitation from 2018 to 2020 for lava domes at Unzen volcano and their relationship to earthquakes, and (2) Haar wavelet analysis to understand the response of displacement to precipitation. The method is based on dome displacement from the Unzen Ground Based Synthetic Aperture Radar system and 48-hour rainfall from the MP radar rain gauge system. As a result, the authors confirmed the following: (1) precipitation of 150 mm or more in 48 hours tends to increase the vertical fluctuation of the dome, and even rainfall of less than 100 mm per 48 hours has a similar effect when it is repeated in an intensive manner; (2) After precipitation, major dome displacement can take days or weeks to occur, and is not instantaneous like the dome collapses in Soufriere and Merapi.


Introduction
Volcanic domes are ubiquitous landforms that are even to be found in areas where effusive volcanism dominates the eruption processes (e.g., Iceland: rhyolitic domes underneath the ice [1], but most domes are associated with andesitic [2] to rhyolitic lava [3].Thus, they are predominantly located along the subduction zones of the Ring of Fire and the American edge of the Pacific from Canada to the Southern tip of Chile (Fig. 1).Scientists have long recognized that lava domes can be the source of hazards to the 1313 (2024) 012026 IOP Publishing doi:10.1088/1755-1315/1313/1/012026 2 surrounding population, the economic and the infrastructure assets (e.g.road marking [4]), because of their gravitational collapse during eruption, triggering pyroclastic-flows, and because of the domes' explosions that can propel ash and scoria high in the upper-atmosphere, and thus affect large areas under the wind as well as international air-traffic.As a rule of thumb, in the aftermath of eruptions, internal pressure at the dome decreases and the risk of dome explosion reduces, even if non-eruptive phreatic and phreatomagmatic explosion can still occur, like it did between 2012 and 2014 at Merapi Volcano [5].Furthermore, volcanic domes tend to grow and "break-down" during eruptions, with only small domes (gravitationally stable) or virtually none remaining in between eruptions.There are however exceptions to this general vision, and volcanoes like the Shiveluch in Kamchatka, which experienced 6 years break in dome growth for instance [6], or like the Unzen Volcano on Shimabara Peninsula in Japan are well known for having generated domes that remained on the slopes and did slide on the volcanic slope under gravity.

Research location
Unzen-Fugendake is an active Volcano located in Southwest Japan on Kyushu Island, where the whole volcanic complex forms the Shimabara Peninsula (Fig. 2).The volcanic complex is enclosed in an intraarc volcanic graben, which slices the Shimabara Peninsula in the East-West direction with an approximate length of 30-40 km, and with an offset of ~200 m [7].Historically, Unzen Fugendake has been active in 1663, then in 1792, with both times lava effusion [8].The volcano last erupted during the period 1990-1995, creating a series of dacite lava-domes, which collapsed and generated series of blockand-ash flow-types pyroclastic flows intercalated with debris flow deposits.Therefore, past eruptions have been dominated by effusive processes and vesiculated pumiceous materials are very rarer [7].Even after the eruption ended at Unzen Fugendake, several lobes of the dome remained imbricated one into another at the summit, falling on the Eastern side of the volcano (Fig. 3).Fumaroles continue to be active to date, with "white smoke" evading the top of the dome.This activity has been linked to an intravolcanic aquifer that is being recharged from rainwater [9].It entails that volcaniclastic sediments underneath the dome are also experiencing wetting and drying periods, which are arguably influencing the stability of the overlying dome.If dome collapses during eruptive phases have been linked to rainfall activity [10], there is a lack of evidence for domes in between eruptive phases.In the present contribution, the authors are examining this theory, to define (1) whether there is dome movement is occurring under gravity; and (2) whether this movement is connected to rainfall activities.

3.1.XRAIN & GBSAR
To reach these two goals, an investigation period of two years' records between 2018 and 2022 was used (timespan constrained by the deployment period of the GBSAR).The rainfall data at the summit of the Unzen Volcano were generated from XRAIN data over a region of 0.91 x 1.61 km (Fig. 2).XRAIN (eXtended RAder Information Network) is a high-performance radar rain gauge network operated by MLIT (Ministry of Land, Infrastructure, Transport and Tourism), which uses X-and Cband MP (multiparameter) radar to provide minute-by-minute observations at 250-meter intervals.The XRAIN data used in this study were 48-hour summations of 10-minute rainfall amounts for the target area obtained from DIAS (Data Integration & Analysis System) (https://diasjp.net).The dome movement was recorded using the GBSAR sensor (Ground Based System Aperture Radar) installed by the MLIT (Ministry of Land, Infrastructure and Tourism) and their local office for Unzen reconstruction.(Fig. 4a).GBSAR has the great advantage of continuously measuring minute displacements of lava domes without the need to reinstall the instrument in inaccessible locations.In this study, data were provided from June 2, 2018, to December 31, 2020, and the data were analysed separately for each of the five dome blocks using data sets collected over a 48-hour return period (Fig. 4b).The displacements from GBSAR are the average of the displacements for each of the observed dome point clouds.In this study, the basis for the displacement is the position of the dome in the 48 hours immediately preceding each point cloud.This position is defined as 0 mm, and the displacement is defined as the distance moved during the 48 hours.

3.2.Haar Wavelet Transform
By considering each dome displacement as a frequency response, the Haar wavelet transform was applied to each dome displacement in this study.By repeatedly applying the Haar wavelet transform to the" approximate components", the overall pattern of displacement is visualized and its relationship to local rainfall is understood.The Haar wavelet transform was applied a total of four times to visualize the relationship between the response patterns of each dome and rainfall.During the period from June 2018 to December 2020, there were at least 10 heavy rainfall events with 48-hour total rainfall exceeding 150 mm, six of which were clearly and independently separated: the first occurred early in the observation period at the end of June 2018, the second at the beginning of July 2019, then at the end of March and end of August, and also occurred at the end of March 2020 and early July and end of July 2020 (Fig. 5).In addition to these events, two other rainfall events occurred repeatedly in the two wettest months, September 2018 (Total: 536 mm) and September 2020 (Total: 683 mm), with total precipitation amounts slightly less than 150 mm (red squares in Fig. 5).Dome displacements (separated by domes 1-5) during the same period show positive and negative vertical displacements in the range of ~4mm to +3mm (Fig. 5).A time series plot of rainfall data versus dome displacement data shows that there is no immediate dome response to rainfall (Fig. 5).

Result
IOP Publishing doi:10.1088/1755-1315/1313/1/0120267 However, each significant rainfall event with rainfall exceeding 150 mm per 48 hours is associated with a delayed peak in dome movement activity (Events 1, 2, and 3 in Fig. 5).The peak intensity of rainfall is accompanied by a peak in dome movement for each of these events over a period of days to weeks (immediate to one week for the peak).Furthermore, the peaks of displacement of different parts of the domes are not the same for any rainfall event that causes any movement.In fact, domes 1 and 2 tend to have relatively little movement and less vertical oscillation, both positive and negative, while domes 3 through 5 experience high activity, increasing in the order of dome number (Fig. 5).In other words, domes located near the summit (domes 1 and 2) are more stable and less prone to movement than domes 3, 4, and 5, which are located lower down the slope.The movement of the domes was also examined in relation to the earthquake observed at Unzen Fugendake, but no clear response to shaking was observed (Fig. 5).

4.2.Haar wavelet analysis
Despite of the different amplitudes of the different domes' movements, thespatial-temporal displacement for each recorded variation is showing similar patterns (Fig. 6).Decomposing the movement of each section of the dome (shown in Fig. 5), into 4 different levels shows several-months scale variations, which may either be artefacts or the result of longer-term processes, such as degassing, movement of the volcanic chimney column inside the volcano, etc.These large-scale movements seem to be relatively regular over the 2 years' study periods.Besides those variations, more localized variations in time correspond to the rainfall-linked movement of the dome (there is no direct physical evidence, it is a statistical relation that can be drawn at this point).The level 1 of wavelet decomposition shows that same-scale variations are occurring for all the domes (Fig. 6), and that these short scales variations are following the same relations of dome 1 and 2 being the less mobile, while domes 3 to 5 are the most mobile (Fig. 6).This differentiation, however, disappears for the other levels of variations.For the wavelet level 2, dome 5 still shows more accentuated negative movements, but this pattern is not apparent for level 3 and 4. The movement that have been temporarily linked to rainfall events > 150 mm are not the only one triggering movements of the dome.The two events 4 and 5 (Fig. 5 and Fig. 6) with prolonged rainfalls over a month period, with peaks between 100 mm and 150 mm are also followed by increased dome activity.
Finally, there is a "hinge" period in March 2019 and May 2020, where a series of high-intensity, shortlived movements are recorded that are not associated with rainfall, but in contrast to other periods of activity not associated with rainfall, this peak of activity can be seen at all wavelet levels and, more importantly, it also appears to be the boundary between the two periods noted to be associated with rainfall (Fig. 6): Before March 2019, wavelet level 4 shows wavy fluctuations, but from this date until September 2020, such movements are rarely seen and return at the end of the study period.

Discussion
The present contribution is showing that the dome of Unzen Volcano is undergoing gravitational activity with both increase and decrease in altitude, and that this movement is not homogeneous for the whole dome, but that the different portions of the dome are behaving differently.The upper part of the dome near the summit seems to be less prone to deformation compared to the lower lobes (domes 3, 4 and 5).This activity can be linked in most cases to heavy rainfalls >150 mm in 48 hours, or consecutive events between 100 mm and 150 mm during a month (the two months of September in this study).The dome is thus behaving more like a landslide would with a time lag between movement and the rainfall due to the time needed for infiltration to occur and internal processes to happen.According to Hoshimizu, an aquifer in the volcanic cone exists, although the author states that it is of low-storage capacity [7].This aquifer has also been shown to recharge with rainfalls, and it is thus likely that the aquifer recharge in the dome may create vertical variations and other geometrical changes at the surface, which are influencing the dome movement (up and down).
At the beginning of 2019, the data shows a peak of activity that does not seem to be directly related to rainfalls, and this variation is recorded in all the different wavelet levels.Moreover, wavelet level 4 shows that there is a pre-Spring 2019 and a post-Spring 2019 in the overall behavior of the entire dome.One possibility behind this movement may be related to seismic activity as a tectonic earthquake occurred in Kumamoto Prefecture on January 3rd, 2019 (33,1.6N/ 130,33.2E)at a depth of 10 km for a Mw.Of 5.1 and an intensity of 6 at the surface in Kumamoto prefecture and an intensity of 3 in Shimabara City at the foot of Unzen Volcano (Source: The Japanese Meteorological Agency).Like for the rainfall events and this earthquake, the link between the dome movements and these events are based on common temporality and present scientific knowledge, but assumptions are difficult to assess further at this stage, and dataset over longer periods as well as geotechnical data are most likely necessary to become more assertive.

Conclusion
The dome of Unzen Volcano moves in the aftermath of rainfall events in excess of 150 mm per 48 hours and in the aftermath of series of events > 100 mm.Background rainfalls also seem to play an accompanying role as single isolated peaks are not linked to important movement.Background mobility of the dome seem also to be controlled by other external elements, some of which might be seismic activity as well, which we argue has the potential to modify groundwater pathways.

Figure 1 .
Figure 1.Map of volcanoes with lava domes with clear eruption records map constructed from the data from the Smithsonian Institution National Museum of Natural History, Volcanism.

Figure 4 .
Figure 4. Location of GBSAR, with the area between the orange lines being the observation range (a); Block division of the lava dome data used in this study (b).

Figure 5 .
Figure 5.Time series observations of the different lobes of the dome, with the precipitations for the observation period of 2018-2020.During the period from June 2018 to December 2020, there were at least 10 heavy rainfall events with 48-hour total rainfall exceeding 150 mm, six of which were clearly and independently separated: the first occurred early in the observation period at the end of June 2018, the second at the beginning of July 2019, then at the end of March and end of August, and also occurred at the end of March 2020 and early July and end of July 2020 (Fig.5).In addition to these events, two other rainfall events occurred repeatedly in the two wettest months, September 2018 (Total: 536 mm) and September 2020 (Total: 683 mm), with total precipitation amounts slightly less than 150 mm (red squares in Fig.5).Dome displacements (separated by domes 1-5) during the same period show positive and negative vertical displacements in the range of ~4mm to +3mm (Fig.5).A time series plot of rainfall data versus dome displacement data shows that there is no immediate dome response to rainfall (Fig.5).

Figure 6 .
Figure 6.Approximate components differentiated by the spatial-temporal amplitude of the dome displacement using the Haar Wavelet transform; (a) one transform; (b) two transforms; (c) three transforms; (d) four transforms.