Effect of Wind

Wind speed and direction were routinely measured at 15-minute intervals during the eruption (fig. 2). The average speed throughout the entire eruption was 4.8 meters per second (m/s). The northeast wind direction, or trade winds, dominated in frequency; the south wind, or Kona winds, was comparatively infrequent. Until July 14, 2011, the wind speed and direction were recorded by an instrument at HVO; thereafter, the data were also recorded from an instrument at North Pit, on the north rim of Halemaʻumaʻu. Both instruments recorded similar wind speeds and directions, although the North Pit instrument typically recorded slower wind speeds (0.1–0.4 m/s) than the HVO instrument, but the readings could be variable. Because of the comparable data, the two wind gages provided a consistent dataset throughout the ten years of the eruption. For the sake of uniform data, figure 2 shows data sourced from the instrument at HVO.

Wind speed was compared with tephra accumulation rate to test the hypothesis that the accumulation rate would vary with wind speed. There was little effect of the wind speed on the average network accumulation rate (fig. 3F). However, the wind speed affected the accumulation rate at individual buckets in an unanticipated way. Buckets A1 and A4, downwind during the trade winds, show a peak in accumulation rate at a wind speed of about 5 m/s (figs. 3A, B). Seemingly, slower speeds are less able to transport tephra (mostly ash) to the buckets, and faster speeds tend to carry the ash past the buckets. Bucket A8A shows a similar pattern, although the peak accumulation rate is lower because of a slightly lower wind speed (fig. 3C). The wind effect was minor or absent at bucket A7 (fig. 3D), which was almost directly above the southeast (SE) sink, an area of downwelling against the southeastern crater wall that remained throughout the eruption. Bucket A7 most likely collected tephra supplied directly by the thermally buoyant plume above the sink before it was advected away by the wind.

Bucket A9, upwind from the spattering sites during trade winds, shows a different pattern (fig. 3E). Many of the low wind speeds were during the Kona winds, so the accumulation rate was highest at such times. The stronger winds were typically trade winds, and the accumulation rate decreased accordingly as increasing speed more efficiently transported ash away from bucket A9.

To our knowledge, this is the first time that such detailed influences of wind speed on transport and deposition of ash have been recognized, at least in a proximal setting. The effects are apparent because of the long duration of the eruption, the presence of local wind gages, and the repeated production of tephra by weak spattering during a variety of wind speeds. However, there is still uncertainty whether stronger fountaining and ash plumes with more convection would have experienced similar wind effects on a medial or distal spatial scale.

Accumulation Rates at Buckets A4 and A7

Throughout the eruption, buckets A4 and A7 typically acquired the most tephra on a day-to-day basis. Figure 4 shows that A4 contained the most tephra more than 860 times, and A7 more than 960 times, totaling more than 1,820 times out of the approximate 2,400 collections; buckets A8, A8A, and A9 accounted for most of the remaining approximate 580 collections. Bucket A4 was directly down the trade wind direction from the south end of the lava lake, and bucket A7 was directly above the south end. Accordingly, it is not surprising that these two buckets generally acquired the most tephra mass.

Buckets A4 and A7 each accounted for most of the collected tephra mass but at different periods during the eruption, likely reflecting changing conditions within the Overlook crater (fig. 4). Until late 2012, the lava lake within the Overlook crater was either absent or very deep and only sporadically visible. Rockfalls were numerous during this time, and trade winds blew mostly lithic ash that fell into bucket A4. Fairly little convective thermal updraft reached the rim of Halemaʻumaʻu when the lake surface was low, so bucket A7, directly above the widening crater, received little ash compared to bucket A4. From late 2012 onward, however, the lake level was sufficiently high enough that spattering could often send juvenile ash into bucket A7, largely carried by the convective thermal updraft, especially above the SE sink. As a result, the daily amount of juvenile mass collected in the network slowly rose in 2013–2015 before increasing by several times in 2016 with rising lake level, particularly accumulating in A7. Tephra fall continued at bucket A4 in 2016, but it was mostly juvenile and much less abundant than at A7. Mintz and others (2021; chap. E of this volume) describe the nature of spattering in early 2015, and such activity characterized the SE sink during the later high stand of the lava lake.

Juvenile Percentage of Tephra During the Eruption

Large changes in tephra componentry with time characterized the decade-long eruption (fig. 5). These changes are defined by the proportions of juvenile components (Pele’s hair and tears, hollow spherules, and bubble walls) and lithic components (clasts derived from the wall of the Overlook crater and secondary alteration minerals, chiefly anhydrite). The changes mainly reflect different levels of lava in the lake and the interaction of the lava column with the wall of the crater.

Starting in 2018, the juvenile component percentage each day was generally greater than 98 percent, whereas all the ash was lithic on some days during the first 4 years of the eruption. The juvenile to lithic proportions ranged widely during the first 2 months of the eruption, reflecting the continued deep level of lava and the repeated rock falls from the overhanging crater wall. The rock falls produced fine debris that fell onto the lake surface and were incorporated into the spattering from the lake.

The juvenile proportion began to gradually increase, from 20 to 50 percent in early June 2008 to 60 to 90 percent by late November 2008. The first view of the lava lake was on September 5, 2008 (Patrick and others, 2021; chap. A of this volume), perhaps indicative of a rising lava level during the summer that was responsible for the increasing juvenile component percentage. The crater widened during this time, contributing more wall-rock clasts to the erupted tephra.

The juvenile to lithic component ratio plummeted starting in early December 2008, as lake level withdrew (Patrick and others, 2021; chap. A of this volume), bottoming at 0 percent on December 8 and remaining less than 10 percent until February 4, 2009. The volume of ash was small during these two months and may have been mainly rock-fall dust wafted from the crater by thermal updrafts.

After this two-month period, the juvenile to lithic ratio gradually increased, and by May–June 2009 it was up to 70 to 90 percent (fig. 5). This period resembles that of summer and fall 2008, as the lake level was probably rising and able to contribute more spatter to the rim of Halemaʻumaʻu. Similar to 2008, this 2009 period ended abruptly on June 30 with withdrawal of the lake, and for the next 4.5 months few juvenile components were collected relative to the previous months, with a juvenile to lithic ratio generally less than 40 percent. The surface of the lava lake was deeper than previously measured, however, and only sporadically increased to higher ratios, most notably on September 17, 2009, when the juvenile to lithic ratio reached nearly 70 percent.

In late 2009, the juvenile to lithic ratio increased again and remained greater than 70 percent until June 2010 before decreasing to 30 to 60 percent for the next 3–4 months. Afterward, the ratio varied widely into 2011, lowering to <10 percent in March during the Kamoamoa fissure eruption (March 5–9, 2011) (Patrick and others, 2021; chap. A of this volume) when the lake level dropped, and it remained low through April. Thereafter, the lake level began a slow rise, and the juvenile to lithic ratio followed suit, with notable reversals, continuing through to the end of the eruption in 2018.

The increase of juvenile components after 2011, particularly in 2012, is striking and had at least three causes. The main cause was rising lava lake level; the development of the consistently high-level lake starting in late 2012 brought sites of spattering closer to the rim of Halemaʻumaʻu, resulting in more juvenile tephra falling into the bucket network (fig. 6). Thus, the juvenile percentage increased without a need to decrease the amount of nonjuvenile tephra in the buckets. This is particularly evident starting in late 2015 but is apparent between 2012 and 2015 as well (fig. 6).

Two other causes of increasing juvenile proportion are minor but still noteworthy. One is that the drowning of the crater wall by rising lake level made less wall material available to supply lithic material to the tephra. This leads to an absolute loss in the lithic components. The other cause is that the crater wall above the lake was gradually stabilizing. For the first several years of the eruption, much of the wall, especially around the northern part of the Overlook crater, was overhanging and, hence, unstable. With time, most of the overhang was lost through rock falls, so there was less lithic material that could fall directly into the lake and be erupted as tephra.

The most noticeable increase in collected juvenile mass took place over about a 2-month period in December 2015 and January 2016 (fig. 6). At that time, lake level rose from a 2-year (2014–15) average elevation of 977 m to a 2-year (2016-17) average of 1,000 m. This 23-meter rise shortened the average depth between the spattering sites and the rim of Halemaʻumaʻu from 131 to 108 m. That 23-meter change accompanied more than a sixfold increase in the average daily collected juvenile mass from 0.7 grams per day (g/d) in 2014–15 to 4.5 g/d in 2015–16. The SE sink was the principal spatter site throughout this four-year period, so that the increase in collected juvenile mass cannot be ascribed to shifting spattering locations. Instead, it must reflect either an increase in the gas content—and therefore explosivity—of the magma or the shorter vertical distance needed for the convective thermal plume above the SE sink to deposit ash and smaller lapilli onto the rim of Halemaʻumaʻu.

Lacking evidence to the contrary, it seems most likely that the 23-meter decrease in vertical distance allowed the convective plume to lift more than six times the amount of juvenile tephra onto the rim and into the collecting buckets. Mueller and others (2018) demonstrated the importance of wall height in governing if tephra left the Kīlauea Iki vent in 1959, where a high wall trapped tephra within the crater. The Halemaʻumaʻu eruption provides a small-scale example of such topographic control on tephra dispersal.

Amount of Tephra Collected

Small amounts of tephra, mostly composed of ash but including rare lapilli (Pele’s tears and small fragments of pumice) rarely more than 10 cm long, were collected from the buckets. For the first eight years of the eruption, it was uncommon for a bucket to contain more than 1 g after 24 hours of deposition. Starting in 2016, however, the tephra accumulation rate was much higher; at least one bucket would contain several grams, and two to three other buckets might have had 1–2 g after 24 hours.

The catchment areas of the buckets, and the amount of tephra that could be collected, varied with time. To facilitate comparisons, a mass per unit area value (m/a), generally expressed as grams per square meter, was calculated from the collected amount in each bucket. All tephra masses and thicknesses in this paper are based on such m/a values unless otherwise noted.

Figure 7A shows the 10-year m/a totals for the ten buckets, not counting fallout from rockfall-induced explosive events. Isomass contours based on m/a values are drawn in the figure. Bucket A7 collected the equivalent of 68.2 kilograms per square meter (kg/m²), nearly double the amount in bucket A4 (35.3 kg/m²). Buckets A8 and A8A collected smaller amounts than bucket A4 despite being nearer the SE sink. Bucket A4 most likely collected ash that rose high enough convectively to have been advected by the trade wind, whereas buckets A8 and A8A received mostly ash that barely rose out of Halemaʻumaʻu. Bucket A7 is favorably located to receive tephra from wind advection and from the lower part of the convection thermal updraft above the spattering sink.

Using an assumed deposit density of 1,300 kilograms per cubic meter (kg/m³) (a common deposit density for fine-grained pyroclastic deposits; Houghton and others, 2013), a cumulative thickness (T₁₃₀₀) of tephra at each bucket site can be calculated (fig. 7B). Values range from about 53 mm at bucket A7 down to 2 mm at bucket A6. Such thicknesses do not match observed total thicknesses, however, because of erosion throughout the 10 years and the fact that Pele’s hair forms a highly inflated deposit with a deposit density much less than 1,300 g/m²; for example, Pele’s hair accumulated around bucket A7 to a measured uncompacted thickness of more than 20 cm during 2016–17.

Thirty-two rockfall-induced explosive events (Orr and others, 2013) deposited bombs or blocks outside Halemaʻumaʻu (table 1). The first and largest event initiated the eruption on March 19, 2008, before the bucket network was installed (Houghton and others, 2011, 2017). Values of m/a for the 31 later explosive events are shown in figure 7C. Bucket A9 had an m/a value of about 53.8 kg/m², almost double the value at bucket A7. This reflects the largest spatter fall of the eruption outside of Halemaʻumaʻu on August 6, 2016; it was directed toward bucket A9, resulting in a measured m/a value of 49.8 kg/m², which is 85 percent of the 10-year total m/a value for that bucket. The 2016 explosion produced a T₁₃₀₀ of 41 mm at bucket A9 (fig. 7D).

Figures 7E and 7F show the 10-year m/a and T₁₃₀₀, respectively, for all the tephra collected in the buckets, including that deposited during explosive events. Bucket A7 netted the most tephra, about 97.4 kg/m² with a T₁₃₀₀ of 75 mm. Bucket A3 acquired the least amount of tephra, about 5.1 kg/m² and a T₁₃₀₀ of 4 mm.

Figure 8 illustrates varying patterns of tephra accumulation. Steps in the curves result from rockfall-induced explosive events. The m/a accumulation rate in 2008 was high at buckets A4 and A7 but not at A9, which was usually upwind of the lake (fig. 8). Starting in 2009, the accumulation rate was faster at A4 than A7 and was rather steady thereafter. The accumulation rate began picking up at A7 in early 2016, as lake level rose and the convection thermal updraft could carry tephra high enough to reach A7. The accumulation rate at A9 began to increase at about the same time but was greatly offset by the explosive event of August 6, 2016.

The m/a accumulation rates at buckets A8 and A8A are broadly similar to one another (fig. 8). The buckets are 7 m apart; A8 is at an azimuth of 75 degrees from A8A and about 3 m farther from the rim of Halemaʻumaʻu. The small difference in location leads to slightly higher accumulation rates at bucket A8A. We originally installed A8A on a fence post 110 cm above ground to test the theory that wind reworking would increase the apparent accumulation rate at A8 (and at all buckets on the ground). Instead, the data show that reworking is a minor concern and that the position of A8A slightly closer to the spattering site controls the difference in accumulation rates. As illustrated by these buckets, tephra accumulation rate in a near-vent setting depends on location.

The amount of tephra that fell into the bucket network probably constituted only a small portion of all the tephra produced during the eruption. A substantial but unquantified amount fell onto the floor of Halemaʻumaʻu, inaccessible to study except for one brief visit by helicopter in 2016. Spattering took place from ephemeral sites around the periphery of the lava lake and frequently deposited tephra on the floor of Halemaʻumaʻu, well below the elevation of the bucket network. For example, the floor of Halemaʻumaʻu was generally less than 60 m above the lake before May 2015 (when it overflowed, it raised the rim by 8 m) and generally less than 30 m after 2016. In contrast, the bucket network was about 85 m above the floor of Halemaʻumaʻu. Consequently, much of the tephra produced by spattering was trapped within Halemaʻumaʻu, most falling back into the lava lake, but some deposited on the floor of Halemaʻumaʻu and on the sloping wall above the southeastern end of the lake.

Because so much tephra was inaccessible, the tephra collected in the buckets is not an accurate gage of the total tephra production from the Overlook crater. It is instead a minimum measure of the output from the SE sink area, which was the most persistent site of downwelling in the lava lake and the most frequent site of spattering in the later years of the eruption. The varying accumulation rate of tephra in the network broadly reflects other observed changes, too; specifically, the level of lava in the lake and subtle changes in elevation of the summit area measured by Global Position System (GPS), both probably resulting from a pulsing magma supply (Swanson and others, 2017).

Distribution of Pele’s Hair

Pele’s hair mainly accumulated between the Overlook crater and the Halemaʻumaʻu parking lot, where it formed a nearly continuous loose blanket and local drifts. It was common to feel Pele’s hair falling on one’s face; the sensation was like walking through a cobweb, and one would invariably try to brush the hair away. First-time visitors in the Halemaʻumaʻu parking lot were always amazed by the drifts of Pele’s hair along the curb, and many found out the hard way that a handful of Pele’s hair left glassy, needlelike shards prickling their hands.

However, Pele’s hair was not confined to the summit area; it was also dispersed far downwind, falling as far as Hawaiian Ocean View Estates, 60 km southwest of the lava lake. The ground surface within several kilometers south of the lava lake had a discontinuous coating of Pele’s hair (fig. 9A), and drifts of loose hair 20–30 cm thick collected downwind of obstacles (fig. 9B, C). With a low sun angle, the surface deposit presented an amber sheen when looking toward the sun, but a dull brown shade when looking away.

The deposits of Pele’s hair formed over the course of the eruption but especially thickened during the final three years when the lake level was high and tephra accumulation rates soared (fig. 10). In addition, some explosive events themselves produced large amounts of Pele’s hair, sometimes creating overnight a pristine fluffy deposit akin to newly fallen snow.

When the southern Kona wind was blowing, Pele’s hair fell north of the caldera in neighboring residential areas, never accumulating into continuous deposits but leaving individual filaments glistening in the sun. Two trees at HVO were sometimes decorated with strands of Pele’s hair 10–20 cm long that resembled tinsel on a Christmas tree.

The presence of Pele’s hair south of the Kīlauea caldera guided the interpretation of a similar deposit that accumulated there in the early 19th century. Subunit L2 of the Keanakākoʻi Tephra (Swanson and Houghton, 2019) is a deposit of hair resembling the modern hair deposits that formed in 2008–2018 and are understood to have had a similar origin from open-vent spattering, likely from a lava lake. In general, any lava lake with persistent spattering will produce abundant Pele’s hair; a recent example at the time of writing is the Masaya lava lake in Nicaragua (Global Volcanism Program, 2025).

Size and Timing of the Explosive Events

Most of the 32 explosive events during the 10-year-long eruption were recognized on HVO webcams and are described in the narrative of the eruption (Patrick and others, 2021; chap. A of this volume). All these events left accumulations of tephra in the buckets that were higher than normal, and most made an identifiable deposit on the ground between buckets (fig. 11). In addition, webcam imagery shows many other rockfall-induced explosive events that were too small to send tephra coarser than medium ash onto the rim of Halemaʻumaʻu.

Separate m/a and isomass maps were prepared for 24 out of 32 explosive events, using bucket accumulations or a combination of those accumulations with m/a measurements of the deposit between and beyond the buckets. The m/a values were contoured, and the bulk volume was estimated from the isomass contours by the method of Pyle (1989) using a deposit density of 1,300 kg/m³. The results indicate that erupted bulk volumes range from 1 cubic meter (m³) to 301 m³, equivalent to values of the volcanic explosivity index (VEI) of −2 to −4 using the expanded VEI scale of Houghton and others (2013), which considers small eruptions not covered by the original VEI scale. No VEI was calculated for eight small explosions because subsequent tephra fall into the buckets contaminated that from the explosions themselves.

Three of the four largest explosions, with VEI of −2, occurred in 2008 (table 1). The first opened the eruption and contained no juvenile components (table 1; Houghton and others, 2011). The other two carried some juvenile clasts, with the September 2, 2008, event dominated by pumiceous lapilli. The fourth VEI −2 explosion took place on May 9, 2018, one day before the lava lake disappeared (fig. 12). The deposit of this explosion could not be studied adequately because of its timing, but it clearly resulted from the largest magmatic explosion of the eruption (table 1).

Most explosions had a juvenile content greater than 60 percent, but five explosions in 2008 and two in 2011 had fewer juvenile components (table 1; fig. 13). These seven exceptions took place when the lava lake was small and deep. Since 2008, the proportion of juvenile to non-juvenile (wall rock and secondary minerals) components showed no consistent change with time. Most of the post-2008 explosive events contained overwhelmingly juvenile components; 13 events had more than 90 percent juvenile material, as determined from componentry observations of tephra in buckets, and 6 of those 13 events had 95 percent or more juvenile material.

With an exception for 2008 and 2016, the explosive events were evenly distributed in time with little clustering (fig. 13). There were, however, eight explosions in 2008 and nine explosions in 2016. Why were 2008 and 2016 so different in terms of explosion timing?

In 2008, the vent in the Overlook crater was small and easily choked by rock falls, which could partly dam escaping magmatic gas. The March 19, 2008 event is a classic example of an explosion driven by magmatic gas but carrying no juvenile clasts (Houghton and others, 2011). Throughout the year, juvenile clasts progressively became more abundant (table 1; figs. 6, 13), but the vent remained small with the lava hidden from view for most of the time.

In contrast to 2008, the lake level was higher in 2016 than it had ever been before. Consequently, the site of explosions was closer to the Halemaʻumaʻu rim than previously, so the ejecta had less distance to travel to reach the rim. In other words, explosions that were weaker than in the past could nonetheless deposit ejecta outside Halemaʻumaʻu. In addition, parts of the crater wall that had previously been affected only slightly when the lake level was low now experienced prolonged heat and acidic gas. The wall responded by frequently shedding rock falls, nine of which resulted in spatter reaching the rim of Halemaʻumaʻu. These rock falls had no discernible influence on the near-daily tephra output before or after them, either in amount or juvenile percentage (fig. 6). The lake level was equally high for most of 2017, but by now the crater wall had equilibrated with the lake, and there were no explosive events because the rock falls were smaller and less frequent.

Some of the explosive events were clustered in time (table 1; fig. 13). Fifteen events took place within six days of one another, and three events happened within eight days in 2015. A likely explanation for the clustering is that a rockfall large enough to trigger one explosion could destabilize other parts of the crater wall, particularly those adjacent to the area of the first fall, leading to subsequent events in a short time span. The observation that 15 of the 32 events occurred within 6 days of one another became useful when assessing future rockfall-induced events.

April 16, 2008, Explosion

An explosion at 0236 Hawaii-Aleutian standard time (HST) on April 16, 2008, differed from all the other explosive events in componentry and dispersal characteristics. It produced a fine pink ash that was advected west-southwest from the Overlook crater, a direction different from the prevalent south-southwest direction (figs. 14, 15; Houghton and others, 2013). An anemometer, an instrument that measures wind speed and direction, at HVO recorded a northeast wind direction of about 50 degrees with a speed of 2.5 m/s at the time of the April 16 eruption. This wind direction was roughly consistent with the dispersal direction of the ash. The dispersed ash avoided most of the bucket network, but its axis was almost exactly along Crater Rim Drive on the Kīlauea caldera floor, so that the deposit could be readily accessed and examined. During daylight on April 16, m/a measurements were made at 26 locations along the paved road, as well as at 13 locations off the road, where the thin ash was difficult to recognize. In a day or two, rain had removed much of the deposit.

The m/a measurements clearly indicated a closed-contour maximum isomass value of about 1,000 grams per square meter (g/m²) located about 1,200 m from the Overlook crater (fig. 14). At that distance, the 100 g/m² isomass contour was only about 150 m from the 1,000 g/m² contour in a direction normal to the dispersal axis. The ash was hard to detect off the pavement, but measurements along the paved road showed that the northern edge of the deposit was about 650 m north of the dispersal axis. Attempts to define the precise edge of the fall deposit were unsuccessful; figure 14 shows locations where we think the ash was absent. The complex isomass contours, with the maximum so far from the vent, made it impossible to apply the method of Pyle (1989) to estimate volume. Consequently, we were not able to calculate a VEI.

The deposit, which was wet when measured, was fine, slippery, and contained few clasts >1 mm in diameter, even in the proximal buckets A1, A4, and A7. The pink color was distinctive; no other explosion during the eruption produced a deposit that was as fine or colorful.

Componentry examination of the April 16 explosion indicated that only about 5 percent of the ash was juvenile. The rest consisted of altered wall rock. The combination of the fine grain size and the altered nature of the ejecta indicates that an area of secondary alteration on the crater wall was the source of most of the ejecta.