The 2019 Eruption Dynamics and Morphology at Ebeko Volcano Monitored by Unoccupied Aircraft Systems (UAS) and Field Stations

: Vulcanian explosions are hazardous and are often spontaneous and direct observations are therefore challenging. Ebeko is an active volcano on Paramushir Island, northern Kuril Islands, showing characteristic Vulcanian-type activity. In 2019, we started a comprehensive survey using a combination of ﬁeld station records and repeated unoccupied aircraft system (UAS) surveys to describe the geomorphological features of the ediﬁce and its evolution during ongoing activity. Seismic data revealed the activity of the volcano and were complemented by monitoring cameras, showing a mean explosion interval of 34 min. Digital terrain data generated from UAS quadcopter photographs allowed for the identiﬁcation of the dimensions of the craters, a structural architecture and the tephra deposition at cm-scale resolution. The UAS was equipped with a thermal camera, which in combination with the terrain data, allowed it to identify fumaroles, volcano-tectonic structures and vents and generate a catalog of 282 thermal spots. The data provide details on a nested crater complex, aligned NNE-SSW, erupting on the northern rim of the former North Crater. Our catalog of thermal spots also follows a similar alignment on the ediﬁce-scale and is also a ﬀ ected by topography on a local scale. This paper provides rare observations at Ebeko volcano and shows details on its Vulcanian eruption style, highlighting the relevance of structural and morphologic control for the geometry of craters and tephra fallout as well as for structurally controlled geothermal activity.


Introduction
Eruptions, their craters and fissures are known to be strongly controlled by pre-existing structures, such as fracture networks [1]. Crustal structures also play an important role during the initiation and lifespan of volcanoes, possibly affecting the locations of magma pathways, hydrothermal fluids and their alteration [2][3][4][5]. The effects of existing crustal structures are identified at the surface, as the morphology of volcanoes evolves in close association with this structural forcing. Also the deposition and erosion of material is controlled by conduit geometry, crater morphology and other structural consequences [6,7].
As deposition of eruptive material often obscures direct views of the underlying structures, general morphological expressions can be indirectly assessed where aligned vents and craters indicate underlying zones of structural weakness [8]. Craters may form in groups and alignments [9], developing 2-pyroxenes. The morphology of Ebeko reaches 1142 m above sea level (as we measured by GNSS), yet the active cones are only approximately 250 m high [40]. The active crater chain of Ebeko (Figure 1d,f) hosts three main craters referred to as the South Crater, Middle Crater and North Crater, with temporally varying cold and hot lakes indicative of a shallow groundwater level and interaction with the geothermal system [46]. Relatively strong eruptions of the Vulcanian type occur episodically, with an activity that appears to migrate along the NNE-SSW Vernadsky Ridge, such as eruptions in 1934- 35,1965, 2011 and 2016-2017 that occurred from the Middle Crater and in 1967-1971, 1987-1991, 2009-2010, 2016-20 from the North Crater and a deep funnel on its northeastern rim [41,47,48]. Sometimes eruptions have even occurred simultaneously at different craters. Starting in April 2018, the eruptions in the active funnel in the North Crater migrated to form a new crater located further westward, which is currently the only active New North Crater. Volcano-tectonic processes, degassing and an increased geothermal gradient are relevant on land and also continue to occur offshore, as inferred from seismic reflection profiles [49].
Remote Sens. 2020, 12, x FOR PEER REVIEW 3 of 21 The majority of erupted products are andesites and basaltic andesites [45] that appear highly porphyritic with 2-pyroxenes. The morphology of Ebeko reaches 1142 m above sea level (as we measured by GNSS), yet the active cones are only approximately 250 m high [40]. The active crater chain of Ebeko (Figure 1d,f) hosts three main craters referred to as the South Crater, Middle Crater and North Crater, with temporally varying cold and hot lakes indicative of a shallow groundwater level and interaction with the geothermal system [46]. Relatively strong eruptions of the Vulcanian type occur episodically, with an activity that appears to migrate along the NNE-SSW Vernadsky Ridge, such as eruptions in 1934- 35,1965, 2011 and 2016-2017 that occurred from the Middle Crater and in 1967-1971, 1987-1991, 2009-2010, 2016-20 from the North Crater and a deep funnel on its northeastern rim [41,47,48]. Sometimes eruptions have even occurred simultaneously at different craters. Starting in April 2018, the eruptions in the active funnel in the North Crater migrated to form a new crater located further westward, which is currently the only active New North Crater. Volcanotectonic processes, degassing and an increased geothermal gradient are relevant on land and also continue to occur offshore, as inferred from seismic reflection profiles [49]. Ebeko volcano is renowned for its intense fumarolic activity and picturesque sulfur towers that cluster in four main fields [41,46], which is also why most previous studies considered the geochemistry and petrology of the volcano, its hydrothermal system and associated alteration products [50,51]. Gas measurements by mobile DOAS systems reveal that 90% of SO2 emissions (approximately 100 t/d) escape from the active North Crater and only 10% or less escape from other fumarole fields [52], suggesting that fluid pathways are channelized by the active conduit system, although the numbers may significantly increase during active periods to an SO2 flux of over 200 t/d Ebeko volcano is renowned for its intense fumarolic activity and picturesque sulfur towers that cluster in four main fields [41,46], which is also why most previous studies considered the geochemistry and petrology of the volcano, its hydrothermal system and associated alteration products [50,51]. Gas measurements by mobile DOAS systems reveal that 90% of SO 2 emissions (approximately 100 t/d) escape from the active North Crater and only 10% or less escape from other fumarole fields [52], suggesting that fluid pathways are channelized by the active conduit system, although the numbers may significantly increase during active periods to an SO 2 flux of over 200 t/d [53]. Selected fumarole fields have been intensely studied, with some showing temperatures exceeding 500 • C [53] and others with conical subterranean structures that connect the fumarole fields at the surface with hydrothermal reservoirs at a depth of 10 m [54] or even 20-40 m [42]. The sites of fumaroles are known to appear and disappear or migrate laterally, sometimes (as in 1983) with the activity shifting by 200 m or more [40]; however, a systematic mapping of the fumarole fields, the crater morphology and fluid pathways has not been conducted yet.
Instrumental and permanent monitoring is very sparse. In 2019, there was one time lapse camera and one seismic station located in western Severo-Kurilsk town at a distance of 7.2 km from the volcano. High levels of natural and artificial noise do not permit the registration of weak seismic activity and the identification of volcanic explosions on seismograms is possible in only a few cases. Previously (in the Soviet Union era), frequent gas sampling was performed by the Institute of Volcanology [40]. Previous studies used thermal survey data and geochemical analysis for precursory and eruption activity assessments [40], which is why modern drone-based and geophysical records will also provide an important basis for robust assessment of the changes at this volcano. Sporadic field surveys allow for the sampling of temperature, gas and water at sites of fluid expulsion [55,56] before the new eruption episode started in October 2016. A modern and comprehensive geophysical characterization of this new activity has not yet been achieved.

Data and Methods
We conducted a 10-day field campaign in July 2019, with daily trips from a base camp (156.0605 • E, 50.6883 • N, 700 m) during which we installed a geophone seismometer, as well as time lapse cameras to record and count the duration of explosions, wind drift and the height of explosions, airfall and associated columns. We conducted repeated UAS overflights on two days with low wind and clear atmosphere conditions. As the UAS carried visual and infrared cameras, the data revealed detailed views of the ground and interior of the crater associated with a major explosion.

Remote Sensing by an Unoccupied Aircraft System (UAS)
UAS surveys were attempted on several days but were only successful on 14 and 15 July 2019 due to clear sky and low wind conditions. The 14 July flights were recorded immediately before and after one of the largest explosions. The 15 July flights were achieved for general mapping purposes and acquiring a dense image overlap for referencing.
UASs were Phantom 4 pro consumer drones equipped with a 20 MPixel optical camera (5472 × 3078 pixels) and a 35 mm focal of 24, set to record images at 0.5 s intervals. The launch site of all UAS flights was at the highest point of Ebeko volcano at 1142 m above sea level (a.s.l.) at E156.0160 • N50.6859 • . The flight height of the surveys was then 300 m above this launchpad at 1442 m a.s.l. All flight paths were preplanned but controlled manually to quickly respond to volcanic explosions, ash fall and sudden wind gusts. The UAS also carried a home-built plate hosting a thermal infrared FLIR Tau2 640 × 512 radiometric camera with a 13 mm lens, recording at wavebands between 7.5 and 13.5 µm, which was linked to a Teax frame grabber and a GPS geotagging system recording at an 8 Hz interval. We defined the calibrated temperature range to 0-500 • C because it was assumed that this range would cover the expected apparent temperature field at Ebeko volcano.
Thermal data depend on various object and environmental properties, which here are mainly the emissivity of the object, the distance, solar reflection, viewing angle and the presence of particles/gases in the electromagnetic radiation path [57] and these properties need to be considered to improve the results [58]. In this study, we assumed a constant emissivity of 0.95, which is typical for volcanic environments [59] and a constant flight height (300 m). Other environmental effects that may influence the derived temperature, such as atmospheric attenuation, were considered using the radiometric correction theme in ThermoViewer software (Teax Technology, vs. 1.3.13), a transmissivity of 0.7, an environmental and path temperature of 10 • C and a humidity of 50%.
In total, we recorded 4800 RGB images using the optical UAS camera and 11,300 thermal infrared images using the FLIR camera, all in the nadir field of view. The image data set was manually controlled for artifacts and cleaned. The thermal image data set was preprocessed in ThermoViewer software and exported in radiometric jpeg form.
We processed all the UAS photogrammetric data (both optical and thermal infrared) using the Structure-from-Motion (SfM) computer vision approach [60]. Using Agisoft Metashape (vs. 1.5.2) software, we achieved point cloud reconstruction on optical and thermal data chunks to reconstruct the three-dimensional morphology, structure and temperature field of the Ebeko volcano summit. Geoinformatics analysis was then performed in ArcGIS (vs. 10.2.1), where we investigated the morphology, identified thermal spots and measured the dimensions and changes in the optical and infrared dataset.
Due to the difficult and hazardous terrain, it was impractical to implement a ground control point (GCP) network and to create reliable georeferencing. The onboard GNSS geotagging that was used was yielding SfM products assumed to be consistent in the within-model horizontal distance measurement errors (0.2-0.3 m) being approximately one order of magnitude lower than the positional error [61]. We associated all models with respect to the best and largest UAS dataset acquired on 14 July 2019 (18:00-18:30 UTC) for accurate and consistent georeferencing by point matching the models. This reference model was based on 900 selected high-quality nadir photographs, yielding 39,000 tie points and a dense cloud consisting of 81 million points used to generate a 19 cm digital elevation model and a 9.8 cm orthomosaic with a total error of 1.9 m. The UAS flights conducted just before and after the explosion were shorter and covered the active crater region only but were conducted at the same heights (300 m above the launch pad), yielding 12,000 tie points, 55 million dense points, DEMs of 20 cm and orthomosaics of 10 cm, respectively. The dataset was compared against the reference dataset, which yielded relative errors that had an RMS of 0.3 m for the optical dataset and 0.5 m for the thermal datasets used in this study.

Seismic Observations
As a seismic unit, we used a 4.5 Hz three-component geophone (3D Geophone PE-6/B manufactured by SENSOR Nederland). A single unit was taken with us in the field at a location (156.023057 • E, 50.694460 • N, 1028 m) very close (600 m) to the center of the active North Crater and was used to identify explosive events. The geophone instrument was buried at 1 m depth, fixed in the ground using three steel spikes and covered by a plastic lid and then by tephra material for insulation. As a data logger, we used a DSS CUBE recorder to save the data with a sampling frequency of 50 Hz and internal batteries as a power supply, as well as a built-in GPS for accurate clock synchronization.
The data were analyzed using the Python package Pyrocko and the Snuffler toolbox [62] to show the waveforms, perform event picking and investigate the spectral density associated with the explosive activity. We bandpassed the data (highpass 0.2, lowpass 2 Hz) and first visually compared alleged explosion events to camera data. Based on this, we define an STA/LTA picker. Short events associated with small rockfalls or earthquakes were not further considered in this study. For spectrogram generation, we considered a window length of 100 s, overlapping 75%, with the Hamming taper function, to produce daily plots, allowing us to obtain an overview of the occurrence of explosions. All timings are given in UTC (local time is UTC+11).

Time-Lapse Camera Observations
We installed time lapse cameras to remotely record the occurrence of explosions, determine wind drift during the explosions and help interpret the other data, such as to distinguish between explosions and outgassing events in the seismic records.
The cameras we used were Reconyx UltraFire trail cameras, which were set to record images every 15 s at a resolution of 2304 × 1296 pixels, 72 dpi. The 5 mm lens has a full frame equivalent focal length of 37 mm. Internal lithium batteries powered the cameras. The time-lapse cameras have no internal GPS clock, which we controlled by synchronization to a GNSS station before the installation and controlled again at the end of the experiment for possible time drift (which were all found to be below the sampling interval and therefore considered nonsignificant). We placed three cameras on small tripods, one camera on the South Crater rim close to the UAS launchpad (156.0137 • E, 50.6852 • N, 1130 m) and the other two other to the east of the active crater, which were frequently covered by ash, which is why we concentrate our analysis on the South Crater rim camera. The camera data were then analyzed using kymograph recording [63,64], which is a sophisticated method of photo sequence analysis. We perform this analysis by considering time lapse camera image data stacks and by choosing a vertical line through the active crater in all images (2880 images in a 12-hr interval) and plotting the RGB pixel values of this line along a time axis. The kymograph is therefore a time-space plot that allows us to quickly identify the occurrence, height and duration of explosions as seen in the camera images ( Figure 2).
Remote Sens. 2020, 12, x FOR PEER REVIEW 6 of 21 small tripods, one camera on the South Crater rim close to the UAS launchpad (156.0137°E, 50.6852°N, 1130 m) and the other two other to the east of the active crater, which were frequently covered by ash, which is why we concentrate our analysis on the South Crater rim camera. The camera data were then analyzed using kymograph recording [63,64], which is a sophisticated method of photo sequence analysis. We perform this analysis by considering time lapse camera image data stacks and by choosing a vertical line through the active crater in all images (2880 images in a 12-hr interval) and plotting the RGB pixel values of this line along a time axis. The kymograph is therefore a time-space plot that allows us to quickly identify the occurrence, height and duration of explosions as seen in the camera images ( Figure 2).

Results
The results are presented in two parts. First, we present the general morphology and structural analysis together with thermal spot density distribution. Second, we describe the records associated with a major explosion.

Results
The results are presented in two parts. First, we present the general morphology and structural analysis together with thermal spot density distribution. Second, we describe the records associated with a major explosion.

General Geomorphology and Structural Analysis
The~9.8 cm resolution orthomosaic reveals details of the structures and, in comparison to the 19 cm DEM, a detailed geomorphological database ( Figure 3). We clearly distinguish the three main craters (the South Crater, Middle Crater hosting a lake and North Crater) and the small new crater in the north referred to as the New North Crater, with areas of 42, 58, 80 and 21 × 10ˆ3 m 2 , respectively.  Figure 6) but are also found at distances greater than 500 m. Some small subsidence pits also formed due to ash-covered snow melting above fumaroles and sites of former lakes (sp in Figure 3), as seen on the western rim of the North Crater or 200 m to the northeast of the active crater (Figure 3b).
The orthomosaic of the thermal camera images reveals the largest anomaly inside the New North Crater, with apparent temperatures reaching 220 • C. We note that this temperature is likely damped by steaming of the crater seen in simultaneous optical images (compare Figure 3a,c). The Middle Crater is not expressed by high temperature and represents the coldest region. Temperature anomalies are found inside the South Crater at two main zones, each with a diameter~50-120 m, but also on the crater flanks ( Figure 3i). In addition, we identify a large number of moderate temperature anomalies exceeding 50 • C on the eastern flank of Ebeko volcano, following the dominant NNE-SSW alignment (Figure 3f). Some of these temperature anomalies have dimensions exceeding 50 m in diameter, representing distinct fumarole fields.
Close comparison of aerial photos, digital terrain models and infrared images allows mapping of the fumaroles, thermal springs and other hot points. We distinguish 282 thermal spots, identified as localized thermal spots surrounded by a cold region. These thermal spots are clustered in four main fields, which are 36 spots in the New North Crater (Figure 3b), 42 spots in the South Crater (Figure 3b,h) and 118 spots on the east flank at the locations of fumaroles and sulfur towers (Figure 3f). The thermal spots identified in the New North Crater reveal the highest temperatures; therefore, the small number of identified spots might be misleading. In addition, we find some isolated anomalies, such as those temporarily exposed in the thick snow cover to the east of the lake-infilled Middle Crater or at the sites of large bombs surrounding the active crater. Locations outside the active craters and the E-flank are rare, with less than 8% of all thermal spots identified. (a) Orthophoto generated from optical UAS camera data using the Structure from Motion (SfM) approach. (b) Thermal spots associated with hot springs and fumaroles identified from thermal cameras and optical data (shown by red dots) and cluster analysis (red shades indicate high clustering density) overlapped on a shaded relief digital terrain model generated from the optical UAS data. sp -subsidence pits. (c) Thermal infrared orthophoto. White boxes show close-ups as illustrated in the center and lowest row of the figure. Note that thermal spots cluster in fields, with a local alignment associated with ravine and ridge topography (panel e). Thermal spots are also located at sites of former eruption vents, as in the South Crater (panel i).
The orthomosaic of the thermal camera images reveals the largest anomaly inside the New North Crater, with apparent temperatures reaching 220 °C. We note that this temperature is likely damped by steaming of the crater seen in simultaneous optical images (compare Figure 3a,c). The Middle Crater is not expressed by high temperature and represents the coldest region. Temperature . Results from the UAS survey. (a) Orthophoto generated from optical UAS camera data using the Structure from Motion (SfM) approach. (b) Thermal spots associated with hot springs and fumaroles identified from thermal cameras and optical data (shown by red dots) and cluster analysis (red shades indicate high clustering density) overlapped on a shaded relief digital terrain model generated from the optical UAS data. sp-subsidence pits. (c) Thermal infrared orthophoto. White boxes show close-ups as illustrated in the center and lowest row of the figure. Note that thermal spots cluster in fields, with a local alignment associated with ravine and ridge topography (panel e). Thermal spots are also located at sites of former eruption vents, as in the South Crater (panel i).

Explosions Observed by Instrumental Network, Geomorphology and Structural Analysis
The instrumental network provided important data for contextualization of the remote sensing observations from UAS data. Explosions exclusively occurred in the New North Crater and no activity was observed in the other craters. Our camera allowed for the visualization of the occurrence, wind drift and scale of the explosions.
The comparison of the camera kymograph with the seismic spectrogram reveals a very good agreement of the retrieved event information during clear sky conditions (Figure 4), with 34 events during daytime (black arrow symbology in Figure 4b). The seismometer allowed the counting of explosion events, which represented 192 events with durations over 5 s, with an average of 34 min intervals between the eruptions for the entire measurement period of 10-14 July 2019. However, a closer look at the records documents very uneven activity, as revealed on the clear sky day of 13 July 2019, when the camera kymograph allows for the confirmation of the occurrence of dark ash-laden clouds. In the first 6 h of daylight, only minor activity was associated with steaming and mild explosions, such as at 20:05 UTC (8:05 local time), drifting westwards. Then, eruptive activity strongly increased, with 11 major events in the second half of the day, the strongest identified by camera at 23:40, 1:15 and 8:29 UTC. The seismic data generally confirm this but also reveal a number of events in the < 10 Hz spectral frequency (although with smaller amplitudes and durations), associated with degassing pulses and white plumes, as observed by the cameras. The between-explosion activity is mainly related to degassing, as observed by the white steam plume escaping from the New North Crater. The strongest steaming occurs immediately after explosions, gradually decreasing until almost virtual absence before the next explosion is initiated. The maximum eruption heights cannot be deduced from the close camera location, but the seismic strength and duration of the signal concurs with the visual strength of the eruption, exceeding the image frame in height (which was set to 500 m, Figure 3c).  The eruption lasted at least 2 min, which was also confirmed in the seismic data ( Figure 5). In fact, the seismogram suggests that the eruption initiates rapidly, possibly in two pulses, with a power spectral density peak at 4 Hz. The spectrogram reveals a > 10 Hz frequency lasting 160 s and a much longer and lower frequency of < 10 Hz lasting approximately 420 s. The comparison of optical and seismic data suggests that the higher frequency is dominated during explosions, followed by a < 10 Hz frequency signal tail associated with gas jets and steaming activity. The eruption lasted at least 2 min, which was also confirmed in the seismic data ( Figure 5). In fact, the seismogram suggests that the eruption initiates rapidly, possibly in two pulses, with a power spectral density peak at 4 Hz. The spectrogram reveals a > 10 Hz frequency lasting 160 s and a much longer and lower frequency of < 10 Hz lasting approximately 420 s. The comparison of optical and seismic data suggests that the higher frequency is dominated during explosions, followed by a < 10 Hz frequency signal tail associated with gas jets and steaming activity. UAS flights conducted in the 20 min immediately before and after the 23:40 eruption indicate important changes ( Figure 6) inside and surrounding the active crater. A comparison of apparent temperature fields reveals that before the explosion, the outer flanks are cold and the inner crater is expressed by apparent temperatures mainly between 120 and 160 °C (Figure 6a,d). The flight conducted immediately after the explosions shows a pronounced deposition field 81,200 m 2 in size, marked by elevated temperature on the outer NE flank of the active crater (Figure 6b), which is in agreement with the prevailing wind direction of N70°W. The temperature-derived deposition zone is composed of a large number of small and distinguishable particles, which were interpreted to be bombs and fine materials (Figure 6b). The temperature decreases with distance to the New North Crater, but the isolines of equal temperature show some pronounced complexities (Figure 6c). Temperature isolines show a strong morphological dependency, such as the inner part of the large North Crater being elevated in temperature (Figure 6c). The temperature field inside the active crater at this time is between 160 and 220 °C for most of the pixels and therefore strongly increased compared to the pre-eruption survey. UAS flights conducted in the 20 min immediately before and after the 23:40 eruption indicate important changes ( Figure 6) inside and surrounding the active crater. A comparison of apparent temperature fields reveals that before the explosion, the outer flanks are cold and the inner crater is expressed by apparent temperatures mainly between 120 and 160 • C (Figure 6a,d). The flight conducted immediately after the explosions shows a pronounced deposition field 81,200 m 2 in size, marked by elevated temperature on the outer NE flank of the active crater (Figure 6b), which is in agreement with the prevailing wind direction of N70 • W. The temperature-derived deposition zone is composed of a large number of small and distinguishable particles, which were interpreted to be bombs and fine materials (Figure 6b). The temperature decreases with distance to the New North Crater, but the isolines of equal temperature show some pronounced complexities (Figure 6c). Temperature isolines show a strong morphological dependency, such as the inner part of the large North Crater being elevated in temperature (Figure 6c). The temperature field inside the active crater at this time is between 160 and 220 • C for most of the pixels and therefore strongly increased compared to the pre-eruption survey. Structural analysis reveals that a clear WNW-ESE temperature gradient is seen in the northern inner slope of the crater. This gradient can be better understood by considering the morphological scars that are oriented WNW-ESE, as shown in the slope map (Figure 6f), implying that the temperature field (and deposition of hot fresh material) is affected by morphology and/or by the fluid pathways changing at the rim of the previous North Crater where this New North Crater has evolved. An interesting finding is the zonation of the temperature fields inside the crater, revealing a generally elevated crater field, a main field with increased temperature and an inner field. The inner field is elongated in the NNE-SSW direction and therefore similar to the arrangements of the main craters and the alignment of temperature anomalies and thermal spots on the flank of Ebeko volcano.

Discussion
Ebeko volcano displays Vulcanian explosive activity at the northern edge of the North Crater, where a new 70-m deep explosive funnel has evolved. Using the UAS data, details of the geomorphology and structures of this New North Crater and its environment were mapped. The digital terrain model together with orthophotos and infrared imagery reveals that the crater formed on the northern continuation of a crater row aligned NNE-SSW and was strongly affected by other pre-existing structures as sketched in Figure 7. We compared the results to in situ geophysical observations and were able to determine (i) the general eruption frequency and NNE-SSW alignment of partially overlapping craters, (ii) the NNE-SSW alignment of temperature anomaly fields, (iii) the expression of ~300 thermal spots, (iv) the geometry of the newly formed crater and its inner temperature anomaly and (v) the deposition of new materials associated with a single explosion. These results are novel for the Ebeko site but may also help to better understand other volcanoes where access and structures are difficult to reach, highlighting the efficient use of UAS together with time lapse and geophone records for temporal monitoring. In the following, we first discuss the Structural analysis reveals that a clear WNW-ESE temperature gradient is seen in the northern inner slope of the crater. This gradient can be better understood by considering the morphological scars that are oriented WNW-ESE, as shown in the slope map (Figure 6f), implying that the temperature field (and deposition of hot fresh material) is affected by morphology and/or by the fluid pathways changing at the rim of the previous North Crater where this New North Crater has evolved. An interesting finding is the zonation of the temperature fields inside the crater, revealing a generally elevated crater field, a main field with increased temperature and an inner field. The inner field is elongated in the NNE-SSW direction and therefore similar to the arrangements of the main craters and the alignment of temperature anomalies and thermal spots on the flank of Ebeko volcano.

Discussion
Ebeko volcano displays Vulcanian explosive activity at the northern edge of the North Crater, where a new 70-m deep explosive funnel has evolved. Using the UAS data, details of the geomorphology and structures of this New North Crater and its environment were mapped. The digital terrain model together with orthophotos and infrared imagery reveals that the crater formed on the northern continuation of a crater row aligned NNE-SSW and was strongly affected by other pre-existing structures as sketched in Figure 7. We compared the results to in situ geophysical observations and were able to determine (i) the general eruption frequency and NNE-SSW alignment of partially overlapping craters, (ii) the NNE-SSW alignment of temperature anomaly fields, (iii) the expression of 300 thermal spots, (iv) the geometry of the newly formed crater and its inner temperature anomaly and (v) the deposition of new materials associated with a single explosion. These results are novel for the Ebeko site but may also help to better understand other volcanoes where access and structures are difficult to reach, highlighting the efficient use of UAS together with time lapse and geophone records for temporal monitoring. In the following, we first discuss the limitations and performance of our approaches and then elaborate on the structural control as well as the wider implications of this work.
Remote Sens. 2020, 12, x FOR PEER REVIEW 13 of 21 limitations and performance of our approaches and then elaborate on the structural control as well as the wider implications of this work.

Limitations and Performance
While the instruments and analysis methods used provide the first detailed accounts of the new eruption episode at Ebeko volcano, a number of limitations are relevant to mention. Most of our analysis is based on UAS data. We used a simple consumer drone with a 30-min flight time, which, when carrying the FLIR Tau2 thermal infrared camera, was reduced to only 18-20 min. Therefore, even though multiple UAS flights were conducted, we could not monitor the initiation of an eruption inside the crater by UAS. Observation of the initiation of an eruption by UAS may shed some light on the surface expression of the conduit geometry. Our UAS flights were conducted before and after eruptions so that we could only approximately interpret the location and dimension of the surface expression of the conduit.
As the UAS flights were conducted during a very high activity phase of the volcano, with tephra and even decimeter-sized bombs found at distances exceeding 500 m, we were not able to record ground control points (GCPs). This limitation comes with increased uncertainties of the absolute position and possible error and artifact propagation. We attempted to control this uncertainty by comparing the SfM point clouds against each other, but cannot rule out any systematic positioning errors. Moreover, the coverage by the UAS, although flying over the most relevant parts of the volcano, was not ideal, as the western flank and the surroundings could not be included in the analysis. This limitation was due to the rapidly changing wind conditions and fog/cloud formation, especially on the western flanks.

Limitations and Performance
While the instruments and analysis methods used provide the first detailed accounts of the new eruption episode at Ebeko volcano, a number of limitations are relevant to mention. Most of our analysis is based on UAS data. We used a simple consumer drone with a 30-min flight time, which, when carrying the FLIR Tau2 thermal infrared camera, was reduced to only 18-20 min. Therefore, even though multiple UAS flights were conducted, we could not monitor the initiation of an eruption inside the crater by UAS. Observation of the initiation of an eruption by UAS may shed some light on the surface expression of the conduit geometry. Our UAS flights were conducted before and after eruptions so that we could only approximately interpret the location and dimension of the surface expression of the conduit.
As the UAS flights were conducted during a very high activity phase of the volcano, with tephra and even decimeter-sized bombs found at distances exceeding 500 m, we were not able to record ground control points (GCPs). This limitation comes with increased uncertainties of the absolute position and possible error and artifact propagation. We attempted to control this uncertainty by comparing the SfM point clouds against each other, but cannot rule out any systematic positioning errors. Moreover, the coverage by the UAS, although flying over the most relevant parts of the volcano, was not ideal, as the western flank and the surroundings could not be included in the analysis. This limitation was due to the rapidly changing wind conditions and fog/cloud formation, especially on the western flanks.
The thermal infrared UAS survey was conducted from a flight height of 300 m above the launch pad (1142 m a.s.l) so that the base of the New North Crater was seen in the UAS camera images from a distance of approximately 400 m. Precise measurements of the apparent temperature strongly depend on distance to the target [65], so the true temperature may be underestimated by our measurements. Independent measurements at fumaroles suggest temperatures up to 490 • C [52] or even exceeding 500 • C in the new explosive funnel in 2015 [53]. In addition, steaming activity may dampen the infrared signal and reduce the apparent temperature, while the strength of the method is that the structural analysis is less affected by such parameters [18,28]. We therefore note that the true temperature might be higher than the measured temperature, although the temperatures of only 160-220 • C in the inner crater as measured immediately (within~20 min) after an explosion were indeed surprising. While we cannot explain these low temperatures, we may speculate that (i) the eruptive products are indeed relatively low temperatures due to cooling by surface water (the crater floor is lower than the nearby cold water level in the maar-like Middle Crater), (ii) fine-grained (possibly remobilized) ash quickly covers the materials and dampens the temperature or (iii) our measurements are biased by measurement artifacts arising from small-scale thermal structures, large measurement distances and steaming effects.
The measurements performed using time lapse cameras and a single geophone instrument gave simple but highly useful information on eruption occurrence, wind drift and eruption strength. The instruments were properly working and were hence used for a first-order characterization of the eruption activity and intervals. The comparison of camera and seismic records indicated that none of the methods alone would have been complete, as we observed strong seismicity during periods of degassing (which by mistake could be interpreted as an explosion), as well as strong nearly horizontal westward drift of eruption plumes (which by mistake might have been missed by the cameras placed further east on the volcano slopes). As no other systematic measurements are accessible for the earlier activity episodes at Ebeko, we cannot compare our results to previous records.

Structural Influence
Using the UAS data, we identified approximately 300 thermal spots. Earlier volcanologic mapping was conducted based on aerial photography but dates back over 30 years (the most comprehensive photogrammetric survey was in 1988) and could distinguish between thermal areas and main fumaroles and pots of thermal waters, counting only 25 such sites [41]. As no thermal infrared camera was available during this time, the work relied on a combination of remote sensing and field validation. However, as previous researchers noted the migration of fumarole sites by as much as 200 m (as in 1983; [40]), our detailed and accessible maps provide a unique basis for further studies on the dynamics of thermal spots and fumaroles. Previous studies reveal that both magmatic and hydrothermal activity is migrating, appearing at one site while ceasing at another site, with possible systematic trends following the NNE-SSW structures.
The eruptive activity at Ebeko migrated back and forth along the NNE-SSW crater row, with explosions from the Middle Crater (1934-35, 1965, 2011 and 2016-2017) and from the North Crater (1967-1971, 1987-1991, 2009-2010, 2016-20) and some of the cases exhibited simultaneous explosions at both craters. In addition to the changing vent positions, a migration of individual fumaroles and thermal fields was also identified. One may therefore conjecture that fluid pathways seal and reopen, with systematics that on a large scale are strongly controlled by the hypothesized NNE-SSW fracture system. Furthermore, we found it puzzling why maar-like lakes occasionally developed along this system, either cold (as currently in the Middle Crater) or hot (as it was in 2005-2007 in the North Crater). Our findings suggest that both magmatic pathways and geothermal pathways are strongly structurally controlled by either fracture sets (e.g., the E-flank thermal field) and/or pre-existing vents (e.g., thermal field in the South Crater) as summarized in Figure 7. A cold water infill in the maar-like Middle Crater may indicate that thermal fluids are not rising there, which could be associated with a currently impermeable system. We may conjecture that the different magmatic and geothermal pathways are either competing with each other, where activity at one venting site causes a decrease in the other venting site or are temporally connected. It is possible that hydrothermal activity changes the permeability, as has been proposed at other locations [66], which may help us to understand the sealing of one vent and the reopening of the other vent. Eye-witness accounts during earlier explosions suggest that the activity occurred simultaneously from two opposing vents at the active funnel of the North Crater, which may support this hypothesis of temporally connected pathways.
Most of the identified thermal spots cluster in fields described earlier (e.g., [46,51,52,54,56]). These fields elongate along E-W ravines but also cluster in a NNE-SSW direction, similar to the arrangements of the nested craters (South, Middle, North Crater and New North crater). We identify a weakly expressed but clearly distinguished zone of thermal fields aligned on the eastern flank of Ebeko. This large-scale (edifice-scale) feature is not similarly seen on smaller scales. Considering any singular thermal field, the distribution of thermal spots appears to be affected by morphology and other local effects and are often aligned subparallel to erosional ravines and ridges, which are east-west on the eastern slope of Ebeko. This observation, a large-scale structural control obscured by morphology and ravines on a smaller scale, is in line with findings made elsewhere [15], highlighting the importance of the scales under consideration. In this view, structural control is dominant on a large edifice scale but obscured on smaller scales by morphology and other effects.

Eruption Dynamics
The eruption dynamics at Ebeko are characterized by Vulcanian eruptions [47] forming bread-crust bombs exceeding 5 m in diameter. Another recent study tested the performance of the atmospheric electric potential gradient, allowing monitoring of the passage of ash plumes over Severo-Kurilsk [67]. Using our seismic and camera records, we could not measure the far field drift but the occurrence of the explosions, as well as the associated near-field ash plume and effects. For the four-day period of 10-14 July 2019, we counted 192 events, yielding an average interval of 34 min between the onset of eruptions. We also note (see Section 4.2), however, a strong irregular activity. On 13 July 2019, in the first 6 h of daylight, only minor activity was associated with steaming and mild explosions, such as at 20:05 UTC (8:05 local time), strongly increasing later that day with 11 major events in the second half of the day (strongest 23:40, 1:15 and 8:29 UTC, respectively). We further find that eruptions often occur in clusters, with double or triple eruptions occurring within a few minutes only, followed by pronounced periods of quiescence. While our short data set and limited number of events do not warrant a more rigorous statistical testing of this idea, the repose gap and clustered explosions might be comparable to the behavior of other (Strombolian type) eruptions at volcanoes [68,69] and geysers [70]. The comparison to Strombolian and geyser explosions is particularly interesting, as the dynamics of Vulcanian-type eruptions also shows periodic behavior and recharge-time-dependent eruption size [71]. A similarity further exists in the presence of double or triple eruptions, i.e., eruptions occurring in very close succession followed by a long repose period (approximately 3 h in our dataset, see also Figure 3c). Therefore, one may speculate that the eruption dynamics at Ebeko are an example of Vulcanian style explosions that show an apparent irregularity modulated by explosions that occur in close succession. Whether this mechanism is controlled from a deep depth by rising fluids, at shallow depths or even at aquifer levels is difficult to ascertain from our available data.

Eruption Deposition and Isopach Analysis
Tephra deposition, in particular, may obscure the structural architecture of volcanoes, but is difficult to monitor and quantify in the field. Analysis of tephra allows us to derive important information on explosive eruptions and intensity and allows indirect assessment of eruption height and dynamics [72]; therefore, parameters on tephra depositions and thickness are critical for the assessment of related volcano hazards [73,74]. In particular, information on the dispersion and distribution of tephra surrounding explosive vents is essential for understanding the volcano [75]; however, data are commonly limited due to poor preservation and lack of access to the deposition zones. As pointed out earlier [76], both the proximal and distal locations are challenging to elaborate in the field, despite the relevance for understanding those deposits. In particular, the proximal region of deposition is complex due to local plume and topographic effects. Characterization of the gradient of deposit thickness and clast size is affected by subjectivity and has remained challenging and rarely preserved due to post-eruption processes, which is why traditional isopach reconstructions are associated with major uncertainties and subjective decisions [77]. Accurate field assessments are highly relevant for developing improved modeling and forecasting techniques [74,76]. As we show, the UAS data may be used for tephra deposit isopach mapping and together with thermal sensors, even allows us to record the associated temperature effects. We demonstrate that the deposition area can be deduced by a temperature anomaly, indicating complex isolines associated with local-scale morphology. We did not attempt to compute the volume of the deposition, which in future works can be regularly inferred from DEM differences under consideration of ground control points.

Final Remark on Hazard Aspects
This study is based on one of the first comprehensive measurements realized for decades at this site and may motivate forthcoming researchers to conduct more dedicated surveys or even implement continuous monitoring. Ebeko is a unique laboratory case for studying volcano-tectonic structures and adds a further destination for the study of Vulcanian style behavior.
Ebeko volcano poses a significant hazard to the local population. A 2500-year reconstruction of the eruptive history of Ebeko [78] showed that all kinds of hazards may be expected; the most relevant hazards for the local population are the effusion of lava flows, tephra falls, passing eruptive plumes and lahars, which is further complemented by significant hazards to the aviation industry [79]. In recent years, and during our stay, the volcano sends gases and tephra dominantly to easterly directions, resulting in gases and tephra falls affecting the town of Severo-Kurilsk. Following a catastrophic tsunami in 1952, the town was relocated, unfortunately now being built on alluvial fans of the main rivers Kuz'minka and Matrosskaya originating near Ebeko volcano, with high lahar risk that can reach the town within 20-30 min without any warning [78]. Therefore, close monitoring of the volcano using a combination of in-situ and remote methods is recommended for early warning.

Conclusions
We achieved a geophysical, geomorphological and structural analysis at Ebeko volcano, a site of intense Vulcanian explosions in 2019. We conducted repeat UAS campaigns and investigated seismic and camera records. The data reveal the occurrence of~200 explosive events during the recording period, with an eruption interval of 34 min on average, although we noted that the explosions and lag times were not regular.
We could generate cm-scale digital terrain models and orthomosaics of the optical and thermal infrared cameras carried by the UAS over the erupting crater. We also found that this new crater follows the general NNE-SSW alignment of the chain of craters that developed since the beginning of the Holocene [80]. The new crater, which initiated as a small explosive funnel on the northern rim of the former North Crater, has now evolved dimensions that are comparable to those of the other craters. The depth of the new crater is even below the water level of the maar-like Middle Crater and is affected by other pre-existing structures. In summary, our analysis ( Figure 7) revealed (i) the general eruption frequency and NNE-SSW alignment of partially overlapping craters, (ii) the NNE-SSW alignment of fumarole thermal fields, which are often obscured at smaller scales by morphology, (iii) the expression of~300 thermal spots, (iv) the elongation of the newly formed crater and its inner temperature anomaly and (v) the deposition of new materials associated with a single explosion. These results are novel for the Ebeko site, but may also help to better understand other volcanoes where access and structures are difficult to reach, highlighting the efficient use of UAS together with time lapse and geophone stations for temporal monitoring.
Author Contributions: Drone data, seismic data and time-lapse camera data were collected by T.R.W., A.B. and M.B. Data processing, SfM, seismic data analysis and kymography by T.R.W. Volcanologic observations and historical analysis by A.B., M.B. and T.K. Dynamic interpretation and discussion contribution by A.A. T.R.W. wrote the manuscript with contributions of all authors. All authors have read and agreed to the published version of the manuscript.
Funding: This is a contribution to VOLCAPSE, a research project funded by the European Research Council under the European Union's H2020 Programme/ERC consolidator grant n. [ERC-CoG 646858].