<?xml-model href='http://www.tei-c.org/release/xml/tei/custom/schema/relaxng/tei_all.rng' schematypens='http://relaxng.org/ns/structure/1.0'?><TEI xmlns="http://www.tei-c.org/ns/1.0">
	<teiHeader>
		<fileDesc>
			<titleStmt><title level='a'>Changes in Crater Morphology Associated With Volcanic Activity at Telica Volcano, Nicaragua</title></titleStmt>
			<publicationStmt>
				<publisher>John Wiley &amp; Sons</publisher>
				<date>07/01/2020</date>
			</publicationStmt>
			<sourceDesc>
				<bibl> 
					<idno type="par_id">10517226</idno>
					<idno type="doi">10.1029/2019GC008889</idno>
					<title level='j'>Geochemistry, Geophysics, Geosystems</title>
<idno>1525-2027</idno>
<biblScope unit="volume">21</biblScope>
<biblScope unit="issue">7</biblScope>					

					<author>Catherine Hanagan</author><author>Peter C La Femina</author><author>Mel Rodgers</author>
				</bibl>
			</sourceDesc>
		</fileDesc>
		<profileDesc>
			<abstract><ab><![CDATA[Volcanic summit craters are typically noted to form by roof collapse into a depressurized magma chamber or by explosive excavation. Recent examples of effusive activity (e.g., Kilauea Volcano, Hawai'i) allowed specifically for quantification of the collapse process. However, small spatiotemporal morphologic change related to background mass wasting and low‐level explosive activity has not been well quantified in volcanic craters. Telica volcano, Nicaragua, is a persistently restless basaltic‐andesite stratovolcano. Telica's persistent restlessness is caused by a long‐lived magmatic‐hydrothermal system with high‐temperature crater fumaroles and low‐frequency seismicity, punctuated by subdecadal, low‐explosivity (VEI 1–2) phreatic eruptions. We use photographic observations (1994 to 2017) and structure‐from‐motion point cloud construction and differencing (2011 to 2017) to analyze changes at Telica in the context of summit crater formation and eruptive precursors. Crater wall retreat (up to 40m) spatially correlates with long‐lived high‐temperature fumaroles in the crater walls, whereas eruptions eject material (>5m) from the crater floor through vent formation and/or clearing. These processes sustain a morphology similar to that of pit craters but without a shallow depressurized magma chamber. Our observations indicate system‐wide sealing prior to eruption by viscous magma in the conduit and eruption of a dome in 2017 and hydrothermal mineralization, not from vent covering talus; though, vent covering talus can redirect the shallow conduit. This study shows promise for photogrammetric techniques in correlating morphologic change with summit crater formation and volcanic activity and the power of long‐term visual observations in understanding active volcanic processes.]]></ab></abstract>
		</profileDesc>
	</teiHeader>
	<text><body xmlns="http://www.tei-c.org/ns/1.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink">
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="1.">Introduction</head><p>Volcanic activity introduces morphologic change to active craters. Large-scale mass movements involving volcanic edifices and craters, such as flank instability and caldera collapse, have been monitored with determinable rates of change. Flank instability at Kilauea (Hawai'i), Piton de la Fournaise (Reunion Island), and Etna (Italy) volcanoes has been monitored with geodetic methods <ref type="bibr">(Poland et al., 2017)</ref>. Caldera formation via roof collapse caused by magma evacuation and depressurization of shallow crustal magma chambers has been recently well-observed and quantified; for example, stopping at Miyakejima volcano (Japan) in 2000 <ref type="bibr">(Geshi et al., 2002)</ref>, trapdoor faulting events at B&#225;r&#240;arbunga volcano (Iceland) in 2014 <ref type="bibr">(Gudmundsson et al., 2016)</ref>, or piecemeal and episodic collapse at Kilauea volcano (Hawai'i) in 2018 <ref type="bibr">(Anderson et al., 2019;</ref><ref type="bibr">Neal et al., 2019)</ref>. However, changes in crater morphology related to frequent or persistent background and low-explosivity activity are not well quantified.</p><p>Background changes within active craters, occurring between main eruptive phases, are often ephemeral and below the detection threshold of common remote sensing techniques. <ref type="bibr">Csatho et al. (2008)</ref> used terrestrial LiDAR data to detect changes (m scale) in the level of the lava lake at Erebus volcano. <ref type="bibr">Walter et al. (2015)</ref> used TerraSAR-X spotlight mode data (m scale) to study the morphology of the lava dome at Merapi volcano. <ref type="bibr">Bagnardi et al. (2016)</ref> used both TanDEM-X radar data and Pleiades-1 satellite imagery to describe topographic change (&lt;m scale) at Fogo volcano from the 2014 to 2015 lava flow. These studies provide high spatial resolution observations of changes in volcanic features associated with effusive eruptive activity but do not investigate changes associated with background activity. Such small-scale changes are noted to contribute to overall crater morphology from strato to shield volcanoes <ref type="bibr">(Harris, 2009;</ref><ref type="bibr">Roche et al., 2001;</ref><ref type="bibr">Rymer et al., 1998)</ref> and help to distinguish between summit pit craters and summit explosion craters. Studies of these processes have relied upon direct visual observation (e.g., <ref type="bibr">Roche et al., 2001)</ref>. Structure-from-motion (SfM) photogrammetry <ref type="bibr">(Westoby et al., 2012)</ref> provides the means to both observe and quantify small-scale intra-crater change through time, presenting an opportunity to analyze crater formation mechanisms and potential precursors to eruptive activity.</p><p>Telica volcano, Nicaragua (Figure <ref type="figure">1</ref>) exhibits both background persistent restlessness and low-level explosive activity contributing to small-scale (i.e., 10 1 -10 2 m and 10 4 -10 5 m 3 ) changes in crater morphology, making it a good case study for summit crater formation mechanisms and correlation to eruptive activity. The morphology of Telica's active crater is characteristic of basaltic shield and intermediate composition stratovolcano calderas or summit pit craters <ref type="bibr">(Geshi et al., 2002;</ref><ref type="bibr">Roche et al., 2001;</ref><ref type="bibr">Simkin &amp; Howard, 1970)</ref>. Although Telica's active crater has been compared to other summit pit craters, such as Masaya volcano (Nicaragua) and Vesuvius volcano (Italy; <ref type="bibr">Roche et al., 2001;</ref><ref type="bibr">van Wyk de Vries, 1993)</ref>, current observations indicate possible formation as a summit explosion crater (e.g., <ref type="bibr">Roche et al., 2001)</ref>. The distinction between the two hinges on the cause of morphologic change, whether it is driven by roof collapse into a depressurized subsurface magma chamber (pit crater) or by explosive activity (explosion crater). This study presents observations and analyses of change in Telica's crater morphology through time caused by persistent background and explosive activity. We analyze change from 1994 to 2017 through field observations, observations from photographs, and observations in published reports. For the period 2011-2017, we use photogrammetry through photorealistic point cloud construction by SfM <ref type="bibr">(Westoby et al., 2012)</ref> and differencing using the Multiscale Model-to-Model Cloud Comparison (M3C2) method <ref type="bibr">(Lague et al., 2013)</ref>. Through correlating change with background and explosive activity, including observations of long-lived fumarolic activity, we provide insights into sustained summit crater formation (i.e., summit explosion versus pit crater) and proposed eruption mechanisms.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="1.1.">Background</head><p>Telica volcano is a persistently restless basaltic to basaltic-andesite stratovolcano in the Maribios Range of western Nicaragua <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Rodgers et al., 2013;</ref><ref type="bibr">Witter et al., 2016)</ref>, part of the northwesttrending Central American Volcanic Arc (Figure <ref type="figure">1a</ref>). The active Holocene crater lies to the west of prehistoric eruptive centers (Figure <ref type="figure">1b</ref>), with the oldest eruptive products in the complex dated to ~330 Kya <ref type="bibr">(Carr et al., 2007)</ref>. The Holocene vent is comprised of two nested craters oriented east-west. The western crater has been breached by the currently active and deeper eastern crater and forms a bench above the active crater (Figures <ref type="figure">1b</ref>,<ref type="figure">c</ref>). The active crater is roughly 200 m deep and 350 m in diameter with steep or oversteepened walls exposing volcanic stratification in the edifice and a near-flat crater floor (Figure <ref type="figure">2</ref>).</p><p>Background persistent activity at Telica includes high rates of seismic unrest, likely driven by a shallow (&lt;2 km), cooling magma body or conduit and an active hydrothermal system <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Rodgers et al., 2013)</ref>. The hydrothermal system is manifest by long-lived, high-temperature fumaroles within the crater walls, new fumaroles formed during vent formation in the active crater, high concentrations and fluxes of magmatic gases <ref type="bibr">(de Moor et al., 2017)</ref>, and hydrothermally altered eruptive products (i.e., ash and ballistics; Figure <ref type="figure">2</ref>). Intermittent periods of incandescence in the vent and high vent temperatures <ref type="bibr">(INETER, 2013;</ref><ref type="bibr">Geirsson et al., 2014)</ref> have been frequently noted following phreatic explosions, again indicating the presence of a shallow (&lt;2 km) magma body.</p><p>Historic eruptive activity at Telica ranges from VEI 1-4, with phreatic VEI 1-2 activity lasting several months characterizing eruptions of the past few decades. The only documented lava flow prior to this study occurred in 1529 <ref type="bibr">(Global Volcanism Program, 2017)</ref>. Thirteen VEI 1-2 phreatic eruptions occurred during the photo collection period from 1994 to 2017 <ref type="bibr">(Global Volcanism Program, 2017)</ref>, ejecting predominantly nonjuvenile material. Two <ref type="bibr">VEI 2 eruptions occurred prior to 2000</ref><ref type="bibr">, on July 31, 1994</ref><ref type="bibr">, and May 21, 1999</ref><ref type="bibr">(Global Volcanism Program, 2017)</ref>. The latter eruptive episode continued into the early 2000s, with column heights up to ~5 km and vent movement from the northern crater wall toward the south of the active crater (Figure <ref type="figure">2</ref>). From 2001 to 2008, seven VEI 1 eruptions occurred with reported ash emission <ref type="bibr">(Global Volcanism Program, 2009)</ref>. The largest VEI 2 eruptive sequence after the 1999-2000 episode lasted from March 7 to June 14, 2011, with maximum column heights of 1.5-2.0 km above the crater rim <ref type="bibr">(Geirsson et al., 2014)</ref>. Another VEI 1 eruption occurred on September 25, 2013, with gas and ash emission <ref type="bibr">(Global Volcanism Program, 2014)</ref>. The May 7, 2015, to May 11, 2016, VEI 2 eruptive sequence consisted of three main eruptive phases <ref type="bibr">(Roman et al., 2019)</ref>. The final phase moved the vent to the southeast and ejected meter size bombs from the crater, killing livestock up to 1 km from the summit. In November 2017, a lava dome was erupted into the middle of the crater, marking a change from decades of amagmatic activity at the surface.</p><p>Telica's discrete phreatic explosions result from failure of a pressurized seal. Two separate mechanisms of sealing have been proposed based on a suite of disparate observations. Based on visual observations of vent infilling and covering by landslide talus, it has been hypothesized that the landslide material seals the upper magmatic-hydrothermal system, resulting in pressurization and eruption <ref type="bibr">(INETER, 2013)</ref>. In contrast, recent multiparameter studies of the 1999 <ref type="bibr">(Rodgers et al., 2013)</ref>, 2011 <ref type="bibr">(Geirsson et al., 2014</ref><ref type="bibr">), and 2015</ref><ref type="bibr">-2016</ref><ref type="bibr">(Roman et al., 2019)</ref> eruptive sequences suggest sealing of the hydrothermal system by mineralization prior to phreatic explosions. Observations made prior to these eruptions and in support of this hypothesis include a decrease to near cessation in low-frequency seismicity, fumarole temperatures, and gas flux, a lack of significant deformation, and the eruption of predominantly nonjuvenile, hydrothermally altered ash. We explore these hypotheses through our photogrammetric study.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head>10.1029/2019GC008889</head><p>Geochemistry, Geophysics, Geosystems 10.1029/2019GC008889</p><p>Geochemistry, Geophysics, Geosystems</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.">Methods</head><p>This study uses: (1) field observations, observations from published reports, and available photos spanning 1994 to 2017;</p><p>(2) SfM analysis of photos from 2011 to 2017; and (3) differential analysis of SfM derived point clouds to describe and quantify the change in crater morphology associated with background persistent activity and VEI 1-2 phreatic explosions.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.1.">Visual Observations From Photographs and Photographic Panoramas</head><p>Observations of Telica's crater include the crater's general appearance, presence of incandescence in the active vent region, description of fumarole locations and activity, and notable mass movements. Photos of the crater have been taken mainly from the same area on the east rim since 1994, with several photo sets from around the entire crater and from drones. Therefore, our visual observations are predominantly of the southern, western, and northern sections of the crater. Persistent degassing from fumaroles often obstructs the view of the crater, so images with extensive volcanic gas did not contribute to the observations and analyses (i.e., other than the fact that degassing was occurring). In addition to the photo data set, previous visual accounts described in INETER Bulletins, Global Volcanism Program Bulletins, and published papers extend the analysis back in time and bridge gaps not covered by our observations and photographic data set. Table <ref type="table">S1</ref> in the supporting information summarizes these observations, and the sources and dates for images are listed in Table <ref type="table">S2</ref>. Representative photos and/or photo panoramas are presented for the period 1994-2017 in Figure <ref type="figure">2</ref>, and larger versions of these photos are in Figure <ref type="figure">S1</ref>.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.2.">SfM Analysis</head><p>Photographs of the active crater were initially taken to document changes to correlate with geophysical and geochemical studies of the volcano (e.g., <ref type="bibr">Geirsson et al., 2014)</ref>. That is, they were not taken with SfM analysis in mind until the later years of our data set. The survey design and observational geometry restrict the results of the SfM to the southern, western, and northern sections of the crater and inner vent. The inaccessibility of the crater made installing ground control points (GCPs) in the crater impossible. We describe the SfM analysis below, followed by a brief description of the main uncertainties in our resulting point clouds and differencing analyses.</p><p>Dense point clouds were constructed in Agisoft Photoscan Professional v1.2 for the following days: August 9 th , 2011, March 21 st , 2013, December 12 th , 2015, March 29 th , 2016, May 14 th , 2016, and November 22 nd , 2017. Point cloud information, including the number and sources of photos used for construction, the number of final points, and general comments, is provided in Table <ref type="table">1</ref>. The dense point clouds for each day are presented in Figure <ref type="figure">S2</ref>. We followed the basic software workflow to create dense point clouds for all but analysis of the May 14 th , 2016 data set. For the May 14 th , 2016 data set, we followed the workflow for photo alignment with GCPs. The sky and flank of the volcano were masked from all images prior to photo alignment to ensure dense point cloud construction of the crater only. We completed photo alignment with generic, prepair selection at high quality, and dense point cloud construction at high quality. Photos with excessive volcanic gas were omitted by trial and error; however, volcanic gas within remaining crater images caused floating points and planes in some point clouds (Table <ref type="table">1</ref>); these spurious points were removed (i.e., cleaned) from the point clouds prior to cloud to cloud alignment and subsequent differential analysis. All point clouds, with the exception of May 2016, were exported with local coordinates in las format. Geochemistry, Geophysics, Geosystems</p><p>We georeferenced the May 2016 point cloud to provide an absolute scale for the cloud comparisons. Photos for this model were taken from four locations on the eastern crater rim (Figure <ref type="figure">S3</ref>). GPS positions were recorded at each location using a WAAS-enabled Garmin GPSX-60c, with accuracy reported at &#177;1 m. At one site, there were issues with the recorded GPS coordinates, so they were not used. Following photo alignment in Agisoft, the camera locations grouped around these coordinates. One photo per location was assigned to an imported place marker in the WGS84 (Agisoft: EPSG::4326) coordinate system. The remaining photos were filtered with the Agisoft tool to their respective markers, georeferencing was updated, and photo alignment was reoptimized to account for the updated GPS information. This point cloud was exported in las format with the WGS84 coordinate system, providing an absolute scale in later point cloud differencing analyses.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.3.">Point Cloud Alignment</head><p>The dense point clouds were imported into the Cloud Compare (CC) software for alignment and differencing (CC Version 2.6.1). The variable image capture locations, number of photos, and environmental conditions during photo collection resulted in inconsistent regions of overlap from one time period to another (Figure <ref type="figure">S2</ref>). Alignment was completed between designated reference and nonreference point clouds, with all clouds connected in the same frame to the georeferenced May 2016 point cloud. For each alignment, we segmented regions of the nonreference point cloud that showed visible change (i.e., we segmented dissimilar regions) to minimize alignment error. Therefore, fine alignment with the built-in CC tool generally included the inactive western crater bench (Figure <ref type="figure">1c</ref>) and/or the upper crater walls  <ref type="table">S3</ref>, and the alignment workflow is pictured in Figure <ref type="figure">S4</ref>.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.4.">Point Cloud Uncertainties</head><p>Following <ref type="bibr">James et al. (2017</ref><ref type="bibr">James et al. ( , 2019))</ref>, we describe the sources of uncertainty in our derived point clouds. These include: (1) survey design and execution, including the geometry of image capture, changes in cameras (i.e., camera errors within and between surveys), camera position uncertainty from uncertainty of GPS positions used for georeferencing, and lack of GCPs across and within the crater; (2) point cloud alignment errors; and</p><p>(3) systematic errors in individual surveys and between surveys, as well as propagated uncertainties into our derived point clouds. The survey design suffers from several issues as described Section 2.1: (1) images were not initially captured with SfM analysis in mind and therefore were collected with different cameras from different locations on the crater rim, and do not have associated camera positions (i.e., GPS locations); ( <ref type="formula">2</ref>) the circumference of the crater is not safely navigable and we limited time on the crater rim; therefore, images were collected from a small arc of an essentially cylindrical crater; and (3) it was not feasible to place GCPs in the crater. With regards to the use of different cameras, Agisoft has different camera models built into its analysis workflow. The nonoptimal survey design could lead to distortion in photo alignment and subsequent derived point clouds. We visually compared our dense point clouds to the photos used (e.g., Figure <ref type="figure">S5</ref>), as a qualitative indication of the distortion (or lack thereof) of the dense point clouds. The ultimate position accuracy, as well as the scale of the resulting dense point clouds, are only as good as the position uncertainty of the GPS unit used, as well as differences in position between the unit and the actual camera location. The GPS position uncertainty is at best &#177;1 m. We add an additional &#177;1 m for the difference in position between the handheld GPS and camera position. We compared the georeferenced point cloud (i.e., May 2016) to a Planet Labs 4 m PlanetScope Scene to check the accuracy and potential distortion of the point cloud. This analysis indicates that the two are comparable to ~5 m, although it was difficult to identify similar geomorphic features in the two data sets. Attempts to compare the May 2016 point cloud with a 10.1029/2019GC008889</p><p>Geochemistry, Geophysics, Geosystems high-resolution DEM (12 m TanDEM-X-derived DEM) was not possible due to distortion of the DEM in the crater region.</p><p>The alignment technique we used should reduce uncertainty by removing areas of change before alignment; alignment errors were &lt;1 m (Table <ref type="table">S3</ref>). Though not ideal, our point clouds align with each other despite different camera models, survey designs, and periods of image capture, lending confidence in and providing an independent check on the accuracy of the SfM workflow presented here. Finally, changes observed and quantified using SfM were correlated with visual observations of change in the crater from raw photos. Based on the above uncertainties and comparisons, we assume that any change greater than 5 m is significant.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.5.">Point Cloud Differencing</head><p>Point cloud differencing provides a quantitative measure of change in crater morphology between photo sampling periods. Differencing was computed following the M3C2 method outlined in <ref type="bibr">Lague et al. (2013)</ref> and implemented as a CC plugin. This approach computes the surface normals of a designated reference point cloud from a set of core points. The normal for each core point is averaged over the points within a "normal scale" of specified diameter about the core point (see figure <ref type="figure">3</ref> of <ref type="bibr">Lague et al., 2013)</ref>. A larger normal scale lessens the impact of local roughness on the orientation of the normal. All points of the reference cloud were designated core points in our analyses. A cylinder oriented along the normal with a chosen "projection scale" diameter determines the average distance between the two compared clouds from the points that fall within the cylinder <ref type="bibr">(Lague et al., 2013)</ref>. The normal scale was set to 10 m for all calculations to minimize the impact of large surface roughness caused by boulders on the crater floor, and the projection scale was set at 5 m. We set the maximum calculation depth to 50 m to detect the maximum change between periods and limit computational run time, then set the registration error for each comparison to the rms error reported from the fine alignment (Table <ref type="table">S3</ref>). Without a direct alignment rms error for the compared clouds in the August 2011 to December 2015 or the August 2011 to November 2017 difference, the rms errors were combined for each alignment that contributed to the difference (i.e., the August 2011-December 2015 rms error was added to the December 2015-May 2016 and May 2016-November 2017 rms errors for the total August 2011 to November 2017 difference; Figure <ref type="figure">S4</ref>) to provide a more realistic registration error. We differenced successive point clouds, designating the older of the two as the reference cloud. This allowed for a time-progressive analysis of change in crater morphology (Figures <ref type="figure">3</ref> and <ref type="figure">4</ref>, Table <ref type="table">S4</ref>). A total difference from 2011 to 2017 was also computed.</p><p>The results of our M3C2 differencing are reported in Table <ref type="table">S4</ref> and shown in Figures <ref type="figure">3</ref> and <ref type="figure">4a</ref> with a color scale in 5 m increments. The differences reported are normal to the surface of the point clouds. Significant change maps were computed with the M3C2 results to provide an estimate of accuracy for detected change (Figure <ref type="figure">S6</ref>). If the distance between the clouds is at or above the 95% confidence interval, which is defined by the registration error, when compared to roughness and point density in that area, the result of the M3C2 calculation at that point is significant (CC Version 2.6.1).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.6.">Point Cloud 2.5D Volume Change</head><p>We computed volume change between successive point clouds with the 2.5D volume calculation tool in CC.</p><p>To minimize any errors introduced during point cloud alignment, only the points of significant change, as calculated with the M3C2 differencing (Figure <ref type="figure">S6</ref>), were differenced with the corresponding later point cloud in time. This yielded estimates of volume loss and addition between significant points of change on the compared point clouds (Figure <ref type="figure">4b</ref> and Table <ref type="table">S4</ref>). These volume changes are minimum estimates, because the incompleteness of our point clouds leaves large regions with no differences. This is true particularly in the vent area, where the largest volume of material accumulates and is extracted. We computed the volume subtraction rate for each period to compare the rate of material loss (Figure <ref type="figure">4b</ref> and Table <ref type="table">S4</ref>).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.">Results</head></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.1.">Visual Observations of the Crater From Photos and Point Clouds</head><p>Distinctive long-lived or recurring features within the active crater include: high-temperature fumaroles (&gt;300&#176;C) in the crater walls and vent area, hydrothermal alteration and mineralization associated with crater and vent fumaroles, incandescence in the vent region, collapse scars in the crater walls, and talus deposits 10.1029/2019GC008889</p><p>Geochemistry, Geophysics, Geosystems on crater floor (Figure <ref type="figure">2</ref>; Table <ref type="table">S1</ref>). Notably, fumaroles on the upper north, northwest, and south crater walls (Figure <ref type="figure">5</ref>) have been observed throughout the study period. The earliest instances of the north and northwest fumaroles within our photo data set occurred in 1994, but images from 1968 show they are longer-lived features (Figure <ref type="figure">5</ref>; <ref type="bibr">Citron, 1968)</ref>. The south fumarole is less frequently documented in photographs but appears in early sketches dating to 1990 <ref type="bibr">(Global Volcanism Program, 1991)</ref>. Hydrothermal mineralization is visible within the crater, appearing on the crater walls and floor with the telltale yellow to white coloration of sulfur, and sulfur and sulfate-mineral deposits; this is particularly evident in 2017 (Figure <ref type="figure">2</ref>). Furthermore, ash and blocks erupted during the 2011 and 2015-2016 explosive activity were dominated by hydrothermal minerals and alteration, respectively <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Roman et al., 2019)</ref>. Crater wall collapse scars correlate spatially with the location of long-lived hightemperature fumaroles. Talus piles are traced to scars on the crater walls, as in December 2015, March 2016, and November 2017 (Figure <ref type="figure">2</ref>). Incandescence often occurs in new vents following VEI 1-2 phreatic explosions, as captured in the August 1999 crater image (Figure <ref type="figure">2</ref>).</p><p>The morphology of the crater changes frequently through time as a result of background mass wasting and low-level explosive eruptions. The largest change observed from 1994 to 2017 was vent migration from the base of the north-northeast crater wall to the south-central crater floor in August 1999 (Figure <ref type="figure">2</ref>; 1999 image). Vent migration again occurred in May 2015. In 2012 and 2013, and prior to the 2015 eruptive episode, ash and talus partially or completely infilled the 2011 vent up to the level of the surrounding crater floor (Figures <ref type="figure">2</ref>, <ref type="figure">3</ref>, and <ref type="figure">6</ref>). In November 2017, an erupted lava dome filled the vent and uplifted a new talus deposit from recent collapse of the northern crater wall (Figures <ref type="figure">2</ref> and <ref type="figure">6</ref>). When not dominated by collapse and ejected blocks, the floor and walls of the crater were covered by ash, deposited from VEI 1-2 phreatic explosions (e.g., May 2015; Figure <ref type="figure">2</ref>). We present the results of our differencing analyses below that quantify these changes.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.2.">M3C2 Differencing</head><p>Our differencing analyses indicate retreat of the crater walls and excavation of the vent region and deposition on the crater floor and in the vent region. From August 9, 2011, to March 21, 2013, we observe ~15 m 10.1029/2019GC008889</p><p>Geochemistry, Geophysics, Geosystems of retreat along the southern crater wall and ~40 m of deposition over the 2011 vent (Figures <ref type="figure">3a</ref> and <ref type="figure">6</ref>). From March 21, 2013, to December 12, 2015, we observed the greatest magnitude of change between any period. Significant retreat of the southeast and west-northwest crater walls occurred, with a maximum of ~25 m from the southeast and ~15 m from the west-northwest walls. Excavation also occurred from the crater floor, with ~40 m of material removed southeast of the 2013 vent location, coincident with the location of the new 2015 vent, indicating southeastward vent migration (Figures <ref type="figure">3b</ref> and <ref type="figure">6</ref>). We observed ~40 m of added material over the 2013 vent, with ~15 m of material also accumulated on the crater floor to the northeast of the vent (Figures <ref type="figure">3b</ref> and <ref type="figure">6</ref>). From December 2015 to March 2016, we observed ~5 m of excavation near the vent (Figures <ref type="figure">3c</ref> and <ref type="figure">6</ref>). From March 2016 to May 2016, little to no change occurred (Figures <ref type="figure">3d</ref> and <ref type="figure">6</ref>). From May 2016 to November 2017, retreat along the northwest wall reached ~15 m and the crater floor was raised by ~35 m by the extrusion of a lava dome (Figures <ref type="figure">3e</ref> and <ref type="figure">6</ref>). We also observed ~5 m high talus deposits on the northwest and east crater floor.</p><p>Total change in crater morphology for our study period is calculated for the long-term difference between the 2011 and 2017 point clouds (Figures <ref type="figure">3f</ref> and <ref type="figure">6</ref>). The M3C2 differencing indicates retreat of the southeast (~40 m) and north to northwest (~20 m) crater walls and maximum deposition of &gt;50 m over the 2011 vent. Overall, widening dominates along the crater walls while deposition occurs near the vent and on the crater floor beneath collapse scars. These results correlate well with our visual observations (previous section).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.2.1.">2.5 D Volume Change</head><p>Results of the 2.5 D volume calculations are presented for each successive point cloud pair in Table <ref type="table">S4</ref> and Figure <ref type="figure">4b</ref>. In this analysis, the negative volume change reported does not differentiate between crater wall retreat or vent excavation. For every period, with the exception of March 2016 to May 2016, volume loss exceeded volume gain. Volume gains and losses fall within a range of 10 3 to 10 5 m 3 (Figure <ref type="figure">4b</ref>; Table <ref type="table">S4</ref>).  <ref type="figure">S6</ref>). These differences are not specific to any location but report the maximum change between any given period associated with the type of activity that occurred during that period.</p><p>10.1029/2019GC008889</p><p>Geochemistry, Geophysics, Geosystems</p><p>The volume loss rate (i.e., the volume of material removed from the active crater walls and floor) during the observational period is between 1.9 &#215; 10 4 and 4.1 &#215; 10 5 m 3 /yr (Figure <ref type="figure">4b</ref>; Table <ref type="table">S4</ref>). The total volume change rate is approximately -2.4 &#215; 10 4 m 3 /yr, taken as the total volume change for 2011 to 2017 over that time window. These volume change rates are considered minima because of the incomplete coverage in our point clouds.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="4.">Discussion</head><p>We discuss our observations and results regarding mass wasting and redistribution of material from the crater walls to the crater floor, ejection of material from the crater associated with VEI 1-2 explosions and formation of new vents, and potential mechanisms driving Telica's persistent activity and frequent phreatic to phreatomagmatic explosions.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="4.1.">Hydrothermal Alteration Promotes Collapse and Redistribution of Material</head><p>Long-lived magmatic, volcanic, and hydrothermal activity weaken volcanic edifices and contribute to increased mass wasting and debris flow susceptibility <ref type="bibr">(Finn et al., 2001;</ref><ref type="bibr">Lopez and Williams, 1993;</ref><ref type="bibr">Reid et al., 2001;</ref><ref type="bibr">Voight &amp; Elsworth, 1997;</ref><ref type="bibr">van Wyk de Vries et al., 2000)</ref>. As part of this process, persistent fumarolic activity also causes hydrothermal alteration and weakening of inner crater wall rock, which we suggest enhances collapse and steepening of crater walls that is characteristic of summit craters and  <ref type="figure">2a</ref>). Note that the 1968 and northern 2017 images also show fumarole activity high on the northern wall (Figure <ref type="figure">2c</ref>), a feature also frequently noted throughout the INETER Bulletin observations. The southernmost, northwestern, and northernmost fumaroles are circled in red and labeled S, NW, and N, respectively.</p><p>10.1029/2019GC008889</p><p>Geochemistry, Geophysics, Geosystems modulates crater widening. Consistent spatial patterns of crater wall collapse during background activity are evident in our visual observations and M3C2 analyses (Figures <ref type="figure">3</ref> and <ref type="figure">4</ref>; Table <ref type="table">S4</ref>). Material is sourced from the crater walls and deposited on the crater floor, sometimes infilling the vent (Figures <ref type="figure">3a</ref> and <ref type="figure">6</ref>). Although collapse can occur anywhere with steep terrain, the observed higher rates of crater wall collapse and mass wasting are spatially correlated with long-lived high-temperature fumaroles found in the crater walls. Crater wall collapse near fumaroles is observed for all time periods except March to May 2016 (Figure <ref type="figure">3a</ref>,b,e,f).</p><p>No explosions occurred between August 2011 and March 2013; therefore, we consider the recorded signal for this period as representative of background mass wasting coincident with typical persistent activity (Figure <ref type="figure">3</ref>). Topographic profiles across the crater at the locations of the southern and northernmost fumaroles show the greatest change in crater width (Figure <ref type="figure">6</ref>). The fumarole high on the northwest wall has been noted as early as 1968 and as recently as 2017 (Figure <ref type="figure">5</ref>). M3C2 differences show more overall mass wasting below the fumarole, corresponding to ~20-25 m of crater wall retreat from 2011 to 2017 (Figures <ref type="figure">3f</ref> and <ref type="figure">6</ref>). In comparison, mass wasting from the crater walls without fumaroles ranges from 0 to 5 m during the same period (Figure <ref type="figure">3f</ref>). Similarly, persistent fumarole activity and collapse coincide in the south, where fumarole activity and collapse were specifically noted in 2006 <ref type="bibr">(Global Volcanism Program, 2009)</ref>, and in the north, where at least 15 m of collapse occurred from 2011 to 2017 (Figures <ref type="figure">2</ref> and <ref type="figure">6</ref>).</p><p>Telica's summit crater has been described as forming by typical pit crater processes: initial roof collapse into a depressurized magma chamber, followed by sustained subsidence of the crater floor and wall collapse along preexisting, outward-dipping faults <ref type="bibr">(Roche et al., 2001)</ref>. In each area of noted persistent fumarole activity in the crater of Telica, collapse occurs haphazardly and not along predefined structures as observed in other summit pit craters. For example, at nearby Masaya volcano (Figure <ref type="figure">1a</ref>), collapse of the Nindiri crater and continued collapse of the Santiago crater is noted along outward dipping faults, structurally related to initial caldera formation <ref type="bibr">(Harris, 2009;</ref><ref type="bibr">Rymer et al., 1998)</ref>. In contrast, preferential collapse at Telica is spatially correlated with high-temperature (&gt;300&#176;C) fumaroles that promote hydrothermal weakening of the wall rock.</p><p>There is no indication of caldera-like faults accommodating this collapse, though fumaroles likely follow fractures in the edifice. Hydrothermally altered ejecta from Telica further support the existence of extensive alteration from this process <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Roman et al., 2019)</ref>. In a morphologic context, this weakening permits collapse of altered material without predefined fault structures in such a way that expands and undercuts the crater walls to sustain summit crater formation and lateral growth (Figures <ref type="figure">3f</ref> and <ref type="figure">6</ref>).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="4.2.">Morphologic Change With Explosive Activity</head><p>Subdecadal, VEI 1-2 phreatic explosions at Telica volcano change the morphology of the crater through the formation of new vents and the eruption of material from the crater floor. These explosions are typically low volume (e.g., ~0.2 &#215; 10 5 m 3 for the 2011 eruptive episode; <ref type="bibr">Geirsson et al., 2014)</ref>, form conical to cylindrical-shaped vents, produce pyroclastic density currents, and can eject ~1 m diameter blocks up to 1 km, posing significant hazard to the surrounding community and tourists who camp within 500 m of the active crater <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Roman et al., 2019)</ref>. Morphologic change in the crater from February 1994 to August 1999 likely resulted from <ref type="bibr">VEI 2 eruptions on July 31, 1994, and</ref><ref type="bibr">May 21, 1999</ref>, and continued activity into the early 2000s. The largest change was the migration of the vent from the north-northeast crater 10.1029/2019GC008889</p><p>Geochemistry, Geophysics, Geosystems wall toward its current location following the May 21, 1999, VEI 2 phreatic eruption (Figure <ref type="figure">2</ref>). In August 1999, the active vent was cylindrical in morphology and had incandescent material at its bottom (Figure <ref type="figure">2</ref>). Seven VEI 1 explosions between August 1999 and November 2009 visibly altered the active crater, widening the 1999 vent (Figure <ref type="figure">2</ref>). Note that although our SfM does not capture the 2011 eruptive sequence, the volume ejection estimates <ref type="bibr">(Geirsson et al., 2014)</ref> are on the same order of magnitude as our estimates for the long-term (2011 to 2017) yearly volume change (~0.24 &#215; 10 5 m 3 /yr; Figure <ref type="figure">4b</ref> and Table <ref type="table">S4</ref>).</p><p>The three main phases of the 2015 to 2016 eruptive sequence significantly changed crater morphology consistent with eruptive activity. Again, the formation of a new vent caused the greatest change (Figures <ref type="figure">3</ref> and <ref type="figure">6</ref>). INETER noted a change in vent location after the <ref type="bibr">May 7, 2015, explosion (Figure 3b;</ref><ref type="bibr">INETER, 2015)</ref>. The new vent was located ~50 m southwest of the 2013 vent and abutting the southern crater wall (Figure <ref type="figure">6</ref>). The second phase of the 2015-2016 eruptive episode, beginning on September 23, 2015, erupted large lithic blocks and produced an ~4 km high ash plume. The most explosive phase began on November 22, 2015, and ejected meter-sized blocks that killed 12 horses and 12 cows up to 1 km from the active crater. This phase of activity widened the vent area further, with encroaching landslide deposits sourced from the southern crater wall (Figure <ref type="figure">2</ref>). The southern crater wall collapsed more between the December 2015 and March 2016 sampling visits to the crater (Figures <ref type="figure">3c</ref> and <ref type="figure">6</ref>) and continued low-level (VEI 1) activity in May 2016 further removed material around the vent, including recent landslide talus (Figures <ref type="figure">3d</ref> and <ref type="figure">6</ref>). Overall, this eruptive sequence was the second case of vent relocation associated with VEI 2 activity and demonstrates that phreatic to phreatomagmatic eruptions can sustain summit crater morphology without pit crater roof collapse (e.g., <ref type="bibr">Roche et al., 2001)</ref>.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="4.3.">Mechanisms Driving Persistent Activity and Phreatic to Phreatomagmatic Explosions</head><p>Persistently restless volcanoes experience long periods of volcanic unrest punctuated by phreatic to magmatic eruptions. The mechanisms that drive persistent activity are not well understood because of a dearth of observations, making eruption forecasts difficult. Telica volcano is persistently restless, where restless phases are punctuated by subdecadal VEI 1-2 phreatic to phreatomagmatic eruptions. Multidisciplinary studies of Telica volcano have improved our understanding of several proposed mechanisms for what drives unrest of a persistently active volcano and eruption triggers <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Rodgers et al., 2013</ref><ref type="bibr">Rodgers et al., , 2015;;</ref><ref type="bibr">Roman et al., 2016</ref><ref type="bibr">Roman et al., , 2019))</ref>. Cooling and degassing of a shallow (&lt;2 km deep) reservoir or conduit of basaltic andesite magma drives the active hydrothermal system, resulting in high background rates of low-frequency seismicity <ref type="bibr">(Rodgers et al., 2013</ref><ref type="bibr">(Rodgers et al., , 2015;;</ref><ref type="bibr">Roman et al., 2016)</ref> and degassing through high-temperature fumaroles in the crater walls and floor (de <ref type="bibr">Moor et al., 2017;</ref><ref type="bibr">Geirsson et al., 2014;</ref><ref type="bibr">Roman et al., 2019)</ref>. Geochemical, petrologic, and geophysical studies of the 2011 and 2015 eruptive episodes suggest sealing of the magmatic-hydrothermal system prior to VEI 1-2 explosive activity <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Roman et al., 2019)</ref>. Furthermore, visual observations, and now our photogrammetric analyses, of the crater show collapse and infill of inner crater vents prior to eruptions, which have been hypothesized as triggers for eruptions <ref type="bibr">(INETER, 2013)</ref>. We explore these mechanisms using our observations and results and propose a new mechanism for persistent activity and eruption triggering.</p><p>Observations of the 2011 and 2015-2016 eruptive sequences indicate sealing at depth supported by hydrothermal minerals present in ash samples, a decrease in fumarole temperatures and SO 2 flux prior to explosions, and cessation of low frequency seismicity <ref type="bibr">(Geirsson et al., 2014;</ref><ref type="bibr">Roman et al., 2016</ref><ref type="bibr">Roman et al., , 2019))</ref>. The observations and analyses presented above do not allow us to directly investigate hydrothermal mineralization and system-wide sealing. We hypothesize that crater fumarolic activity is a proxy for changes in the magma-volcano-hydrothermal system, where a lack of fumarolic activity indicates blocked or sealed hydrothermal pathways. The interaction of shallow meteoric water with heated bedrock complicates this hypothesis because water vapor may persist in fumarole locations despite deeper sealing. Nonetheless, absence of fumarolic activity altogether would suggest system-wide sealing. Our visual observations from May 2015 to 2016 show an absence of fumarolic activity suggestive of deeper sealing of the system (Figure <ref type="figure">2</ref>). This occurred after the initiation of the 2015-2016 eruptive sequence.</p><p>Our observations and analyses do lend themselves to investigating the hypothesis that shallow blocking of the vent by landslide talus seals the shallow hydrothermal system, causing pressure increase and explosion. A similar process has been described for explosive activity at nearby Masaya volcano, where collapse talus blocked vents in the active Santiago crater prior to phreatic explosions <ref type="bibr">(Rymer et al., 1998)</ref>. Images of the 10.1029/2019GC008889 Geochemistry, Geophysics, Geosystems crater prior to the September 2013 eruptions (Figure <ref type="figure">2</ref>) and our SfM (Figure <ref type="figure">6</ref>) and M3C2 analyses (Figure <ref type="figure">3a</ref>) indicate filling of the 2011 vent with ~25 m of collapse talus. INETER personnel proposed the talus-blocked conduit as the cause of pressure build-up for the September 1, 2013, VEI 1 eruption <ref type="bibr">(INETER, 2013)</ref>. We observe that many of the high-temperature crater wall fumaroles were active. If meteoric water interaction was not significant at this time, this suggests no deep sealing of the hydrothermal system and therefore no deeper pressure buildup prior to eruption (Figure <ref type="figure">2</ref>). Following 2013, the eruptive vent subsequently infilled, and the vent shifted southwestward during the 2015 eruptive episode. Assuming a lithostatic load, 25 m of landslide talus with a density of 1,600 kg/m 3 would exert a pressure of ~0.4 MPa on the conduit. This is enough pressure to close fractures in the shallow hydrothermal system (e.g., <ref type="bibr">Taisne &amp; Jaupart, 2008</ref>, and references therein) and therefore could redirect the conduit in the shallow (&lt;500 m) subsurface. We propose that infilling of the vent may provide sufficient pressure to seal the shallowest part of the system and redirect the upper conduit.</p><p>Prior to our observations, the last recorded magmatic eruption of Telica volcano was in 1529. Between May 2016 and November 2017, a small volume lava dome (~70,000 m 3 ) erupted through the existing vent, uplifting new talus deposits on the crater floor (Figures <ref type="figure">2</ref>, <ref type="figure">3e</ref>, and <ref type="figure">6</ref>). No explosions were reported with this event. Ground deformation, seismicity, and a small percentage of glass particles associated with the 2015-2016 eruptive episode suggest migration of magma in the conduit and eruption of juvenile material <ref type="bibr">(Roman et al., 2019)</ref>. The lava dome represents the final stage of viscous magma migrating up the conduit. This viscous magma and its migration are a likely cause of Telica's persistent activity and low explosivity eruptions. As the magma ascends, it provides the heat engine and flux of gases for the hydrothermal system, which are recorded as low frequency seismicity <ref type="bibr">(Rodgers et al., 2013</ref><ref type="bibr">(Rodgers et al., , 2015))</ref>. The viscous nature of the magmas (i.e., crystal rich basaltic andesites; <ref type="bibr">Witter et al., 2016)</ref> and degassing allow for periodic sealing of gas pathways, combined with hydrothermal mineralization, leading to system-wide pressure buildup and eruption (e.g., <ref type="bibr">Fischer et al., 1994;</ref><ref type="bibr">Heap et al., 2019;</ref><ref type="bibr">Taisne &amp; Jaupart, 2008)</ref>.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="5.">Conclusions</head><p>Nonexplosive, background persistent restlessness and periodic (VEI 1-2) explosive activity subject Telica's active crater to frequent morphologic change, contributing to summit explosion crater formation. We utilize photogrammetry to observe and calculate change related to both types of activity from 1994 to 2017. Photo observations and quantitative M3C2 differences (2011 to 2017) reveal a trend of material deposition on the crater floor and over the active vent of &gt;50 m and retreat of the crater walls of up to 40 m (Figure <ref type="figure">3</ref>). Within this pattern, preferential retreat occurs near long-lived, high-temperature fumaroles in the crater wall. Excavation from the crater floor exclusively occurs with eruptive activity and includes new vent formation within the active crater (e.g., from the <ref type="bibr">May 21, 1999, and</ref><ref type="bibr">May 7, 2015, phreatic to phreato-magmatic explosions)</ref>. Overall, this results in a net loss of material on the order of 10 4 m 3 /yr, similar to estimates of a single eruptive episode (Figure <ref type="figure">4b</ref> and Table <ref type="table">S4</ref>). We therefore envision a model for Telica and other explosion craters, whereby the mechanism of sustained crater formation is related to hydrothermal alteration, weakening of wall rock, and subsequent collapse, in concert with excavation from explosions. This model holds true on the decadal timescale, in which no significant edifice change (such as flank collapse) or change in eruptive style occurs.</p><p>Our observations support widespread hydrothermal system sealing as the primary mechanism of pressurebuild-up prior to eruption, as opposed to primary blocking by vent-covering landslide talusdeposits. Although vent-covering talus may not sufficiently seal the whole hydrothermal system for pressure-buildup with explosive potential, it remains an ejecta hazard, can redirect the shallow conduit, and could contribute to the widespread hydrothermal system sealing supported by pre-eruptive observations in <ref type="bibr">Geirsson et al. (2014)</ref> and <ref type="bibr">Roman et al. (2016)</ref>. System sealing is now thought to be related to the migration of a viscous magma rising through the conduit. Future analyses would benefit from more frequent observation and more complete, georeferenced photo data sets that include the eastern crater wall. Multidisciplinary observations of the vent and hydrothermal system and studies of future eruptions and eruptive products should be considered coincidentally to provide the best possible understanding of Telica's eruption processes and hazards. This is particularly true in light of recent dome formation that indicates Telica's altered magmatic state.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head>10.1029/2019GC008889</head><p>Geochemistry, Geophysics, Geosystems</p></div><note xmlns="http://www.tei-c.org/ns/1.0" place="foot" xml:id="foot_0"><p>15252027, 2020, 7, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2019GC008889 by Pennsylvania State University, Wiley Online Library on [25/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License</p></note>
		</body>
		</text>
</TEI>
