ABSTRACT Emplacement models for voluminous sheet flows of the Columbia River flood basalts vary significantly in style and duration, with the latter ranging from as little as one week to decades and even centuries. Testing the efficacy of such models requires detailed field studies and close examination of each stratigraphic unit. The Steens Basalt, the oldest formation of the Columbia River flood basalts, differs from the later formations in that it is composed of stacked successions of thin, commonly inflated flow lobes combined into thicker compound flows, or flow fields. These flow lobes are of limited geographic extent, with relatively high emplacement rates, but they are otherwise similar to modern examples. Evidence for flow inflation in the much larger sheet flows of the Grande Ronde Basalt, Wanapum Basalt, and Saddle Mountains Basalt is also apparent, but with more variable rates of emplacement. For example, the Asotin and Umatilla Members (Saddle Mountains Basalt) and Sentinel Bluffs Member flows (Grande Ronde Basalt) erupted distinct compositions along their linear vent systems, but over 200 km west of their vents, these flows are no longer distinct. Instead, they exist as compositional zones of a single, moderately mixed lava flow. Such flows must have been emplaced rapidly, in perhaps weeks to months, while others have been shown to erupt over much longer time periods. We conclude that emplacement rates may be quite variable throughout the Columbia River flood basalt province, with thin flow units of Steens Basalt erupting continuously and rapidly, and larger inflated sheet flows erupting over variable time spans, some from a few weeks to months, and others over a duration of years.

ABSTRACT The mid-Miocene Strawberry volcanic field of northeastern Oregon is an example of intracontinental flood volcanism that produced lavas of both tholeiitic and calcalkaline compositions derived by open-system processes. Until now, these dominantly calc-alkaline lavas have not been considered to have a petrogenetic origin similar to that of the flood basalts of the Pacific Northwest because of their calc-alkaline composition. These lavas are situated in between and co-erupted with the dominant volcanic field of the Columbia River Basalt Group (CRBG). Due to the timing, location, and diversity of these erupted units, the Strawberry Volcanics may hold valuable information about the role of crustal modification during large magmatic events such as hotspot volcanism. The earliest eruptions of the Strawberry Volcanics began at 16.2 Ma and appear continuous to 15.3 Ma, characterized by low-silica rhyolite. High-silica, A-type rhyolite eruptions followed at 15.3 Ma. The silicic eruptions continued until 14.6 Ma, with an estimated total volume up to ~100 km 3 . The first eruptions of the intermediate lava flows occurred at 15.6 Ma and continued with both tholeiitic and calc-alkaline, and transitional, lavas until 12.5 Ma. Volume estimates of the intermediate lavas are ~1100 km 3 . The mafic lavas are sparse (~2% of total volume) and are distributed throughout the upper sequences, and they appear to be near last to arrive at the surface. Herein, we show that the Strawberry Volcanics are not only related in time and space to the Columbia River Basalt, but they also share some chemical traits, specifically to the Steens Basalt. Evidence of this similarity includes: overlapping normalized incompatible trace-element patterns, selected trace-element ratios, and radiogenic isotopes. Furthermore, we compared the Strawberry rhyolites to the other mid-Miocene rhyolites of eastern Oregon associated with the inception of the Yellowstone–Snake River Plain hotspot and found overlapping eruption ages, trace and rare earth element compositions, and “A-type” rhyolite characteristics. This research concludes that the Strawberry Volcanics were part of the regional basalt to rhyolite magmatism of the Yellowstone–Snake River Plain hotspot.

ABSTRACT The Keanakāko‘i Tephra was deposited from 1500 to ca. 1820 CE, when Kīlauea’s magmatic output was ~2% of the average output during historical times (post–1823 CE). The tephra consists of deposits from numerous phreatomagmatic and phreatic eruptions, three episodes of high lava fountains, and one lava. Fresh glass is available from most tephra units. Major elements and trace elements were determined for glass from 49 tephra units and three pretephra lavas. Olivine crystals from 11 high-MgO tephra glasses were also analyzed. These results were compared to compositions from Kīlauea’s historical period to evaluate ~500 yr of Kīlauea geochemical evolution. Keanakāko‘i Tephra glass composition ranged widely (e.g., 3.4–11.2 wt% MgO). The observed large variations in FeO, CaO, TiO 2 , and K 2 O at a given MgO indicate numerous compositionally distinct parental magmas, with the two early nineteenth-century pumice eruptions showing the most diverse compositions. These two magmas were erupted on opposite sides of the caldera and probably tapped different magma bodies. The common occurrence of high-MgO olivine compositions (forsterite [Fo] 88%–89%) in MgO-rich tephra glasses indicates that primitive magma (Mg# 73–74) was routinely supplied to Kīlauea’s summit. Wide ranges and reverse zoning in olivine core compositions from some units show that magma mixing occurred before some eruptions. Modeling of compositional variations within Keanakāko‘i Tephra units using alphaMELTS showed that the most consistent trends for crystal fractionation involved shallow magma (1–2 km), with low water content (0.2 wt% in parental magma) and oxygen fugacity just below the quartz-fayalite-magnetite (QFM) buffer (–0.5 log units). Keanakāko‘i Tephra glasses have lower La/Yb and Nb/Y ratios than historical Kīlauea lavas. Low ratios have been observed during periods of high magma output for historical lava, which is inconsistent with the low magma output at Kīlauea’s summit during 1500–1820 CE. The most likely explanation for this inconsistency is endogenous growth within Kīlauea during this period, following formation of the modern summit caldera. No correlation was found between glass chemistry and eruption style for Keanakāko‘i Tephra deposits. Glass samples from many explosive units have lower Nb/Y and La/Yb ratios compared to glass from high lava-fountain units and historical effusive eruptions. The explosive character of Keanakāko‘i Tephra eruptions was probably caused by interaction of magma with shallow or surface water.

ABSTRACT The deeply eroded Goat Rocks volcanic complex was a major locus of andesitic volcanism in the Cascade arc in southwest Washington during the late Pliocene to Pleistocene. This volcanic complex includes the remnants of multiple andesitic edifices over an area of ~200 km 2 , centered ~35 km north of Mount Adams on the arc axis. New 40 Ar/ 39 Ar ages for seven samples and U/Pb zircon ages for nine samples indicate a 2.5–2.9 m.y. eruptive history at Goat Rocks. Four eruptive stages are delineated: Tieton Peak (potentially 3.0–2.6 Ma), Bear Creek Mountain (>1.6–1.3 Ma), Lake Creek (1.1–0.6 Ma), and Old Snowy Mountain (0.4–0.1 Ma), each named for the major vent that was active during that time. Lake Creek volcano was the most voluminous of these edifices and probably rose at least 3400 m above sea level with a volume of ~60 km 3 , comparable to nearby active composite volcanoes. Thirty new bulk composition X-ray fluorescence (XRF) and inductively coupled plasma–mass spectrometry analyses from the volcanic complex are presented, in addition to 54 previously unpublished XRF analyses for samples collected by Don Swanson. The compositional variability is greatest in the early and late stages, ranging from basaltic andesite to rhyolite, whereas the more voluminous middle stages are dominated by andesite to dacite. The middle eruptive stages are interpreted to have been a time of peak thermal energy with a mature subvolcanic plexus. In addition, compositions shift from high-K to medium-K compositions with time, which mimics variation across the arc; early eruptive products are similar in composition to those of Mount Adams, and Old Snowy Mountain stage compositions are more similar to those of Mount St. Helens. The life cycle of Goat Rocks volcanic complex provides new perspective on the longevity and evolution of major arc volcanoes, and on the complex distribution of magma in the Cascade arc at the latitudes of southern Washington and adjacent Oregon.

ABSTRACT Volcanoes that interact with the cryosphere preserve indicators of their eruption environments. These glaciovolcanoes and their deposits have powerful potential as proxies of local and global paleoclimates. The Garibaldi volcanic belt is the northern (Canadian) segment of the Cascade volcanic arc. In this study, we compiled a comprehensive database of Quaternary volcanic landforms and deposits in the Garibaldi volcanic belt. We found that the region exhibits a high degree of volcanic diversity, and a significant component of this diversity is due to the abundance of glaciovolcanoes. These include: tuyas, tindars, subglacial tephra cones, ice-impounded lavas, subglacial domes and breccias, subglacial lava flows, and lava-dominated tuyas. As a group, they inform the presence, thickness, and transient properties of ancient, continental-scale ice sheets (i.e., the Cordilleran ice sheet) that have waxed and waned in thickness and extent across the region. We ascribe much of the character of glaciovolcanism in the Garibaldi volcanic belt to a wide range of magma compositions (alkaline basalt to rhyolite) and to the extreme relief of the landscape. We used forensic volcanologic evidence, in conjunction with our database, to define a terrestrial-based reconstruction of ice-sheet thickness and extent that spans the latter half of the Quaternary (i.e., past ~1 m.y.). We then compared our reconstruction to the marine isotope stage (MIS) record and found a number of positive correlations and discordances. We show glaciovolcanoes to be an excellent, and underutilized, proxy for Earth’s paleoclimate, and a powerful tool for reconstructing ice sheets predating the last glaciation.

ABSTRACT Small shield volcanoes with basal diameters <20 km represent the most abundant type of volcano on Venus. These shield volcanoes number >>10 6 in population and often occur in clusters known as shield fields, which have been interpreted to be analogous to basaltic volcanic fields on Earth. Despite previous work on shield fields, questions related to edifice morphology and magma viscosity, timing relations of events across an individual field, volume of erupted material, and the role of tectonic structures are still unresolved. Here, we address those questions through geologic mapping, volumetric calculations, and statistical analysis of possible edifice alignments in six venusian shield fields: Asherat Colles, Chernava Colles, Monoshi Tholus, Nordenflycht Patera, Ran Colles, and Urutonga Colles. Our results indicate that all of these shield fields and their associated deposits are younger than the surrounding units within the mapping areas, and each field displays overlapping temporal relations with local extensional and contractional structures. Each field also displays a lack of a consistent pattern in the temporal distribution of volcanism with regards to edifice type. Analyses of possible edifice alignments suggest edifice trends that are consistent with mapped tectonic structures within all studied fields except Asherat Colles. Comparison of these six venusian fields to terrestrial basaltic volcanic fields shows that venusian fields may be up to two to three orders of magnitude larger in their areal expanse and volume of erupted material. Our results are consistent with previous interpretations of venusian shield fields representing low rates (likely <5 × 10 −4 km 3 /yr) of magma supply feeding these magmatic centers and highlight the effects of the resolution limit of the Magellan data set on interpreting fundamental geologic processes on the venusian surface.

ABSTRACT Intralava geochemical variations resulting from subtle changes in magma composition are used here to provide insights into the spatial-temporal development of large basalt lava flow fields. Recognition that flood basalt lavas are emplaced by inflation processes, akin to modern pāhoehoe lava, provides a spatial and temporal frame-work, both vertically at single locations and laterally between locations, in which to examine lava flow emplacement and lava flow field development. Assuming the lava inflation model, we combined detailed field mapping with analysis of compositional profiles across a single flow field to determine the internal spatio-temporal development of the Palouse Falls flow field, a lava produced by an individual Columbia River flood basalt eruption. Geochemical analyses of samples from constituent lobes of the Palouse Falls lava field demonstrate that systematic compositional whole-rock variations can be traced throughout the flow field from the area of the vent to the distal limits. Chemical heterogeneity within individual lava lobes (and outcrops) shows an increase from lava crusts to cores, e.g., MgO = 3.24–4.23 wt%, Fe 2 O 3 = 14.71–16.05 wt%, Cr = 29–52 ppm, and TiO 2 = 2.83–3.14 wt%. This is accompanied by a decrease in incompatible elements, e.g., Y = 46.1–43.4 ppm, Zr = 207–172 ppm, and V = 397–367 ppm. Systematic compositional variations from the source to distal areas are observed through constituent lobes of the Palouse Falls flow field. However, compositional heterogeneity in any one lobe appears less variable in the middle of the flow field as compared to more proximal and distal margins. Excursions from the general progressive trend from vent to distal limits are also observed and may reflect lateral spread of the flow field during emplacement, resulting in the juxtaposition of lobes of different composition. Transport of magma through connected sheet lobe cores, acting as internal flow pathways to reach the flow front, is interpreted as the method of lava transport. Additionally, this can explain the general paucity of lava tubes within flood basalt provinces. In general, flow field development by a network of lava lobes may account for the occurrence of compositionally similar glasses noted at the proximal and distal ends of some flood basalt lavas.

ABSTRACT The Tieton andesite lavas of the south-central Washington Cascades have lengths of 74 km and 52 km, ranking them among the longest known andesite flows in the world. These two Pleistocene intracanyon flows occupy ancestral canyons of the Tieton River and its tributaries that drained the Goat Rocks volcanic complex, and part of the Naches River valley from its confluence with the Tieton River to Cowiche Creek. Don Swanson was the first to identify the existence of two Tieton andesite flows near Rimrock Lake in his Ph.D. thesis. Our geologic mapping, supported by geochemistry and age dating, confirms the existence of two Tieton andesite flows separated by ~250,000 yr there and in Tieton Canyon. Tieton andesite lavas are high-potassium, calc-alkaline pyroxene andesite to trachyandesite and have excellent columnar jointing, easily viewed along U.S. Highway 12 in Tieton Canyon. The older, longer Tieton andesite flow, Qta 1 has ~62 wt% SiO 2 and a 40 Ar/ 39 Ar date of 1.64 ± 0.07 Ma. It traveled from the east flank of Bear Creek Mountain to Cowiche Canyon. The shorter, younger flow, Qta 2 , has ~60 wt% SiO 2 and a 40 Ar/ 39 Ar date of 1.39 ± 0.10 Ma. It reached only the Oak Creek Wildlife Area Headquarters from its vent on the ridge north of the summit of Bear Creek Mountain. The volumes of Qta 1 and Qta 2 are estimated to be at least 6.6 km 3 and 2.5 km 3 , respectively. A relatively high effusion rate, estimated at 11–18 m 3 /s from a large volume of available magma, combined with confinement in preexisting stream channels contributed to the development of these long andesite flows. Emplacement by flow inflation is suggested based on similarities in physical characteristics observed in Columbia River basalt flows. Emplacement of the Tieton andesite flows lasted on the order of a decade, based on a constant effusion rate. The flows were erupted from Bear Creek Mountain, ~6 km north of the core of the Goat Rocks volcanic complex, and they are not from Black Thumb, as previously inferred; Black Thumb is a mafic dacite plug that is chemically and age distinct from Tieton andesite. The vents at Bear Creek Mountain are characterized by oxidized andesitic agglomeratic debris formed around remnant fins and plugs of a lava dome complex. The composition of the rocks there matches Tieton andesite well. Other andesite flows in the area were likely erupted from the same vent as Qta 2 and partially buried the vent associated with Qta 1 . Olivine basalt, Qob 1 , and basaltic andesite, Qob 2 , flows are in contact with or adjacent to Tieton andesite at several localities and appear to occupy valleys cut in or alongside Tieton andesite lavas. Qob 1 is older than Qta 2 based on field relationships between the flows, although their radiometric ages are indistinguishable (Qob 1 at 1.38 ± 0.25 Ma vs. Qta 2 at 1.39 ± 0.10 Ma). Qob 2 (1.31 ± 0.04 Ma) may have a similar length as that for Qta 2 , suggested by an exposure of Qob 2 recently identified near the confluence of the Naches River and Tieton River. In Tieton Canyon and at other localities near Rimrock Lake, the older flow, Qta 1 , is perched higher on the slope than Qta 2 , indicating a period of erosion and downcutting between emplacements of the two flows. At two locations where the flows are in close proximity to one another, the base of the upper, older flow is ~40 m above the base of the lower flow. Comparisons of paleodrainage stream gradients, as estimated from the elevation of the distal to proximal flow bases of Qta 1 , Qta 2 , and Qob 2 , with those of the modern-day Naches and Tieton Rivers suggest uplift or tilting of the Cascades of almost 7 m since emplacement of the first Tieton andesite flow.

ABSTRACT The Keanakāko‘i Tephra offers an exceptional window into the explosive portion of Kīlauea’s recent past. Once thought to be the products of a single eruption, the deposits instead formed through a wide range of pyroclastic activity during an ~300 yr period following the collapse of the modern caldera in ca. 1500 CE. No single shallow conduit or vent system prevailed during this period, and most of the deposits are confined to distinct sectors around the caldera. Vent position shifted abruptly and repeatedly throughout this time period. This combination of circumstances, influenced by prevailing wind direction, led to rapid lateral changes in the stratigraphy. We define and describe 12 units, several of which are subdivided into subunits or beds, and place them in a framework that reflects volcanologic processes as well as temporal succession. Eruption style and intensity are exceptionally diverse for a basaltic shield volcano. Bulk tephra volumes range from 10 6 to 10 7 m 3 , and the volcanic explosivity index (VEI) ranges from 1 to 3. The most intense activity included high Hawaiian fountaining eruptions, probably associated with caldera-confined lava flows, and subplinian and phreatoplinian explosions. There was also a wide range of less intense phreatomagmatic activity characterized by different magma/water ratios, with products ranging from ballistic block falls, to cross-bedded pyroclastic density current deposits, to fine-grained ash falls commonly bearing accretionary lapilli. Resumption of a Keanakāko‘i style and pattern of volcanism, which seems possible given events unfolding in May–July 2018, has serious implication in terms of future volcanic risk. The hazards associated with every style of explosive activity at Kīlauea summit are quite distinct from the dominantly effusive style of the past 200 yr and should be factored into any future evaluation of risk.

ABSTRACT A series of coarse-grained, relatively well-sorted, but wall rock–rich pyroclastic deposits within Unit H of the Keanakāko‘i deposits at Kīlauea Volcano, Hawai‘i, is the focus of this study. These “ c ” subunits within Unit H consist of alternations between very coarse and relatively well-sorted pyroclastic fall deposits and products of relatively concentrated pyroclastic density currents. They are associated with both accretionary lapilli–bearing ash falls ( a beds) and cross-bedded, fine-grained pyroclastic density current deposits ( b beds). The Unit H sequence is related to phreatomagmatic explosions from multiple sources in the modern caldera, and we infer that most vents for the c subunits were located near the southern part of the caldera. The c beds contain varying proportions of dense, outgassed juvenile bombs and hydrothermally altered wall rock that suggest, along with coarser grain size and good sorting, that fragmentation conditions were relatively dry for phreatomagmatic eruptions and were perhaps aided by the release of magmatic gases from a deep magma source. The c fall subunits, with thinning half distances of 200–300 m, are more widely dispersed than both the most powerful Hawaiian fountaining eruptions and the well-documented historical explosive eruptions at Kīlauea, with proximal dispersal rates similar to historical subplinian eruptions at other volcanoes. The c pyroclastic density currents were erosive and of a style that represents a threat that is underrated at Kīlauea.

ABSTRACT The golden pumice deposit (unit K1) represents one of the latest episodes of Hawaiian fountaining in the Keanakāko‘i Tephra and is the product of the first high fountaining eruption at Kīlauea summit in ~300 yr, since the caldera formed in ca. 1500 CE. We present a new physical characterization of the deposit based on over 200 field sites, all affected by severe erosion, alteration, and silicic encrusting. We detail the deposit geometry, stratigraphic and structural relationships, and componentry to constrain its volume and reconstruct the eruptive sequence. The deposit is then discussed and set against other young episodes of high fountaining at Kīlauea. We interpret the golden pumice as the product of a days-long eruptive sequence with a source located inside a caldera much deeper than that of today. The eruption probably started along a NE-SW–oriented fissure and migrated toward a single vent in the southwestern part of the caldera, where at least two high Hawaiian-style fountains produced a tephra deposit of ~6 × 10 6 m 3 . Stratigraphic contacts reveal that erosion occurred not only between, but also during the fountaining episodes, suggesting heavy rainfall during deposition. Field observations during this study also led to the discovery of the first stratigraphic evidence that the eastern pumice postdates the golden pumice, which contributes to the new definition of the stratigraphy of the Keanakāko‘i Tephra presented in this volume.

ABSTRACT Two small scoria vents were discovered in the Koa‘e fault system, an extensional regime connecting the east and southwest rift zones of Kīlauea that was previously considered to be noneruptive. The chemical composition of the scoria suggests an early to middle nineteenth-century age. The vents prove that magma can intrude several kilometers into the central part of the Koa‘e fault system from the nearest rift zone, supporting previous seismic and geodetic inferences of intrusions into the Koa‘e fault system in the twentieth century. Geodetic studies for the past 50 yr document widening of the Koa‘e fault system at a time-averaged rate of ~4.5 cm/yr, involving mostly coseismic strains, but also creep and displacement related to dike intrusions. These rates are consistent with a longer-term widening rate for the past ~700 yr calculated from crack widths in a lava flow of about that age. The Koa‘e fault system blends into, and is a structural continuation of, the east rift zone. We interpret the locus of intrusion in the east rift zone to have migrated ~6.5 km SE during the past 100,000–125,000 yr, as estimated from linear extrapolation of measured displacement rates across the Koa‘e fault system and east rift zone. The inception of migration is consistent with the onset of the tholeiitic stage at Kīlauea as interpreted by previous studies. As the rift zone moved away from the summit, a marked curvature in the transport pathway developed in order for the rift zone to maintain its connection to the summit magma reservoir. The migration resulted in development of the SE-trending east rift connector, a term we prefer instead of the upper east rift zone. The connector supplies magma to the ENE-trending rift zone from the summit storage complex but is not itself the site of significant magma storage or eruption. The Koa‘e fault system merges into the southwest rift zone, which has been migrating southeastward for an uncertain period of time. Some magma that enters it passes from the summit reservoir complex through the southwest rift connector (seismic southwest rift zone), analogous to the east rift connector. Both connectors reflect the response of magma-transport pathways to asymmetric volcano spreading away from a relatively fixed summit magma reservoir. The ENE structural grain of the Koa‘e fault system and east rift zone pervades Kīlauea’s entire edifice. Most eruptions take place along this trend. The major exception is the southwest rift zone, which may reflect the stresses of Mauna Loa spreading and the Ka‘ōiki fault system. The dominant ENE grain emphasizes the importance of SSE-directed volcano spreading in controlling most of Kīlauea’s tectonic and eruptive behavior.

ABSTRACT Meticulous field observations are a common underpinning of two landmark studies conducted by Don Swanson dealing with the rate at which magma is supplied to Kīlauea Volcano, Hawai‘i. The first combined effusion rate and ground deformation observations to show that the supply rate to Kīlauea was constant at ~0.11 km 3 /yr during three sustained eruptions from 1952 to 1971, a quiescent period at neighboring Mauna Loa volcano. This rate was also interpreted as the steady supply rate from the mantle to both volcanoes combined throughout historical time. The second breakthrough involved field evidence that activity at Kīlauea alternates between dominantly effusive and explosive styles over time scales of several centuries, and that the magma supply rate during explosive periods is only 1%–2% of the rate during effusive periods. For the historical period, several later studies concluded that the supply rate to Kīlauea has varied by as much as an order of magnitude, contrary to Swanson’s earlier suggestion. All such estimates are fraught with uncertainty, given the poorly known amount of magma stored within the volcano’s rift zones as a function of time—an enduring problem and active research topic. Nonetheless, Swanson’s original work remains an important touchstone that spurred many subsequent investigations and refinements. For example, there is strong evidence that Kīlauea experienced a surge in magma supply during 2003–2007 that exceeded the historical average by as much as a factor of two, and that the surge was followed by a comparable lull before the supply rate returned to “normal” by 2016. There is also evidence for supply-rate variations of similar magnitude during the latter part of the twentieth century and possibly earlier, subject to the aforementioned uncertainty in rift-zone storage. The extent to which variations in the magma supply to Kīlauea can be attributed to partitioning between Kīlauea and Mauna Loa, a long-debated topic, remains uncertain. Since Kīlauea’s inception, the net magma supply to the volcano (and also to Lō‘ihi Seamount, since it began growing) has increased, while Mauna Loa’s growth rate has slowed, suggesting that the volcanoes compete for the same magma supply. However, geochemical differences between lavas erupted at Kīlauea and Mauna Loa indicate that they do not share a homogeneous mantle source or common lithospheric magma plumbing system. Both ideas might be correct; i.e., Kīlauea and Mauna Loa magmas may be sourced in differing portions of the same melt accumulation zone and ascend through different crustal pathways, but those pathways interact through stress or pressure changes that modulate the supply to each volcano. Currently, magma supply-rate estimates are facilitated by comprehensive imaging of surface deformation and topographic change coupled with measurements of gas emissions. Physics-based models are being developed within a probabilistic framework to provide rigorous estimates of model parameters, including magma supply rate, and their uncertainties. Further refinement will require intensive multiparameter observations of the entire magmatic system—from source to surface and above, and from the volcanoes’ summits to their submerged lower flanks—in order to account fully for a complex magma budget.

ABSTRACT Characterization of the subsurface structure of a volcanic edifice is essential to understanding volcanic behavior. One of the best-studied volcanoes is Kīlauea (Island of Hawai‘i). Geological evidence suggests that the formation of the summit caldera of Kīlauea is cyclic, with repeated collapse followed by filling with lava. The most recent collapse occurred ca. 1500 CE, producing a basin that is several hundred meters deeper than the current caldera. In this study, we used two- and three-dimensional gravity modeling of spatially dense gravity data covering the summit area to suggest that, since its formation in 1500 CE, the caldera has been progressively filled by lava flows that are slightly denser than those found in the rim and outboard of the caldera. The geometry of this fill, inferred from gravity data, enables us to reconstruct the morphology of the 1500 CE caldera before its subsequent filling. The coincidence of fumarolic zones and thermal anomalies observed at the surface with the interpreted 1500 CE caldera rim suggests that hydrothermal fluid circulation is guided by the more permeable inner faults bounding the main caldera.

ABSTRACT Kīlauea Volcano’s active summit lava lake posed hazards to downwind residents and over 1.6 million Hawai‘i Volcanoes National Park visitors each year during 2008–2018. The lava lake surface was dynamic; crustal plates separated by incandescent cracks moved across the lake as magma circulated below. We hypothesize that these dynamic thermal patterns were related to changes in other volcanic processes, such that sequences of thermal images may provide information about eruption parameters that are sometimes difficult to measure. The ability to learn about concurrent gas emissions and seismic activity from a remote thermal time-lapse camera would be beneficial when conditions are too hazardous for field measurements. We applied a machine learning algorithm called self-organizing maps (SOM) to thermal infrared time-lapse images of the lava lake collected hourly over 23 April–21 October 2013 ( n = 4354). The SOM algorithm can take thousands of seemingly different images, each representing the spatial distribution of relative temperature across the lava lake surface, and group them into clusters based on their similarities. We then related the resulting clusters to sulfur dioxide emissions and seismic tremor activity to characterizeties between the SOM classification and different emplacement conditions. The SOM classification results are highly sensitive to the normalization method applied to the input images. The standard pixel-by-pixel normalization method yields a cluster of images defined by the highest observed SO 2 emission levels, elevated surface temperatures, and a high proportion of cracks between crustal plates. When lava lake surface patterns are isolated by minimizing the effect of temperature variation between images, relationships with seismic tremor activity emerge, revealing an “intense spatter” cluster, characterized by unstable, broken-up crustal plate patterns on the lava lake surface. This proof of concept study provides a basis for extending the SOM classification method to hazard forecasting and real-time volcanic monitoring applications, as well as comparative studies at other lava lakes.