Characterising the wind-advected medial fall deposit from fissure 8 fountaining during the 2018 lower East Rift Zone eruption, Kīlauea

The 2018 eruption on the lower East Rift Zone, Hawaii, involved the opening of 24 fissures before the eruption focussed on a single point source, fissure 8 (F8). This study characterises the preserved medial F8 tephra deposit using an isopach map, maximum clast size data, and total grain size distribution analysis, shedding light on the tephra transport and dispersal mechanisms beyond the F8 cone occurring during the fountaining. The medial sheet-like deposit covers approximately 0.22 km2, best fit by a Power-Law thinning rate. The TephraFits model estimated the corresponding volume of the continuous medial tephra blanket to be ~ 2 ×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 104 m3, just 0.02% of the total volume erupted from fissure 8. Samples from the preserved medial deposit have grain size modes of − 3.5 to − 4 Φ, compatible with Voronoi tessellation calculations. Maximum clast size did not show a ‘typical’ fining relationship with distance from the vent; instead, it shows no clear pattern. One factor was that the extremely low clast density, a function of a secondary vesiculation event, enabled the pyroclasts to be re-entrained, often repeatedly, by large eddies downwind of the vent. This should be considered in future studies of prolonged fountaining episodes as the clasts involved in the medial fall are rarely well preserved in the geologic record due to their fragile nature but their presence adds complexity to the inferred eruption dynamics.


Introduction
In 2018, an extended period of low (~ 80 m) Hawaiian fountaining occurred within the lower East Rift Zone (LERZ), Hawaii.Activity soon focussed on a single fissure, fissure 8 (F8, Fig. 1), that was situated in the middle of a residential area, known as Leilani Estates.The eruption created a 13-km-long lava flow field (Dietterich et al. 2021) and a new ~ 40-m-high scoria cone, comprised of course-grained tephra, both causing extensive damage to properties and infrastructure (Meredith et al. 2022).Given the sparsity of Hawaiian fountaining studies and the level of risk that Hawaiian fountain eruptions pose to society, understanding the volume, grain size distribution, and eruption dynamics is vital to accurately assess the risk of future eruptions.
Tephra is fragmented material ejected during an explosive eruption and can be further classified in three size ranges: bombs (> 64 mm), lapilli (2-64 mm), and ash (< 2 mm).The tephra of Hawaiian fountaining events commonly contains Pele's hair and other glassy fragments which can cause eye and skin irritation.This along with other hazards associated with these events, such as respiratory issues (Horwell et al. 2008;Longo et al. 2010), a reduction in agricultural productivity, and long-term property and infrastructural damage (Longo 2013), makes field characterisation and modelling of tephra dispersal an important field of research (Wilson et al. 2012;Jenkins et al. 2015Jenkins et al. , 2018)).
Tephra deposits can be characterised by isopach and isopleth maps, vesicularity, and total grain size distribution (TGSD); each shedding light on the dynamics of tephra production and dispersal.In the last decade, statistical toolboxes have been developed to estimate eruption volumes (Biass et al. 2019a(Biass et al. , 2019b) ) and to integrate grain size data obtained in the field to produce a TGSD (Biass and Bonadonna 2014).Based on the errors associated with field measurements and the compilation of isopach maps (as described by Klawonn et al. 2014a, b andKlawonn et al. 2014b), the toolboxes inbuilt error analysis, using Monte Carlo simulations, provide a 5 and 95% confidence interval on the volume estimations.
Whilst the number of tephra studies has increased sharply over the last few decades, only a handful of studies have investigated the tephra dispersal associated with Hawaiian fountaining events (e.g.Andronico et al. 2008;Klawonn et al. 2014aKlawonn et al. , 2014b;;Mueller et al. 2018) and even fewer include a TGSD.Several reasons might account for this gap: (a) Hawaiian fountaining products often having a small dispersal area (a few square kilometres; Parfitt and Wilson 1999).This then requires scientists to enter dangerous or inaccessible proximal localities in order to obtain a collection of representative samples.Alongside this, many Hawaiian fountains are unsteady, with episodes of varied intensity, adding an extra level of difficulty when reconstructing often necessitating episodes of contrasting intensity and dispersal within an eruption being grouped together (Klawonn et al. 2014a(Klawonn et al. , 2014b)); (b) even soon after an eruption, deposits are generally poorly exposed (Mueller et al. 2019) and/or eroded by wind or human activity.Within this study, we present isopach and isopleth maps alongside a TGSD analysis of the medial tephra deposit of the main, powerful phase of the fissure 8 LERZ eruption, illustrating how aerodynamics can add complexity to the inferred eruption dynamics.The opportunity to study such a fresh deposit and subsequently continually assess its characteristics in the future will provide insight into the preservation of Hawaiian fountaining deposits.This step forward in our understanding will be of great use to the community when studying older Hawaiian fountaining eruptions.

Hawaiian fountaining
Hawaiian eruption alongside Strombolian eruptions are the most common, yet weakest, explosive volcanic eruptions to occur globally (Houghton et al. 2016).Lava fountaining can be grouped, based on height, with fountains > 400 m considered high, 100-400 moderate, and any fountain < 100 m considered low fountaining (Houghton et al. 2016).The vast majority of lava fountain eruptions at Kīlauea and worldwide are low (Houghton et al. 2016;Edwards et al. 2020) yet the associated lavas can be highly destructive.The main phase from fissure 8 during the 2018 LERZ eruption was exceptionally long-lived, lasting 93 days (Neal et al., 2019) and accounting for ~ 92-96% of the approximately 1 km 3 of erupted lava (Gansecki et al 2019).The maximum fountain height was only ~ 80 m (Lyons et al. 2021).Whilst heights provide insight into the pyroclast exit velocities and the volatile content, they do not constrain the erupted mass as this is dependent on numerous other variables including the duration of eruption and geometry of the erupting vent.The highest recorded fountaining event, Episode 16 of the 1959 Kīlauea Iki eruption, produced around 10 10 kg of magma over a 3-h period (Richter et al. 1970), but other eruptions with low fountaining have lasted for a significantly longer duration and hence produced similar or larger erupted masses.Typically Hawaiian fountaining events have durations on the order of 10 3 -10 6 s (~ 16 min to ~ 12 days) with some events, e.g.Kīlauea between 1983 and 1986, lasting for ~ 10 7 s (~ 120 days).The main F8 episode in 2018 at the LERZ was on the high end of the temporal range lasting for 93 days (Neal et al. 2019).
Products of Hawaiian fountains are mostly basaltic in composition and are initially formed by primary fragmentation by stretching and inertia of fluid magma (Gonnermann 2015).Namiki and Manga (2008) suggest that the viscosity must be < 10 6 Pa s for fragmentation to occur via this method.When the magma is erupted during fountaining, the initially molten pyroclasts cool by heat exchange with the surrounding gas and air.If the pyroclasts cool below the glass transition, they land as solid particles known as either scoria or golden pumice (lower and higher vesicularity respectively).If the time in the air is too short for the majority of a clast to cool past the glass transition, it will land as a molten fragment known as spatter.Low fountaining events or pyroclasts deposited close to the vent therefore have a lower and shorter flight and hence cooling time and often produce welded spatter deposits.A range of processes superimpose secondary fragmentation (i.e.fragmentation of pre-existing pyroclasts) on basaltic explosive eruptions, thus reducing the grain size of the final deposit.These include ductile failure of clasts falling from a lava fountain (Walker and Houghton 2019), breakup during particle-particle collisions in fountains or jets (Taddeucci et al. 2017), renewed vesiculation and brittle fragmentation of bombs (Namiki et al. 2021), and impact on landing (Richter et al. 1970).

Lower East Rift Zone 2018 eruption
Eruptive activity of Kīlauea in early 2018 was concentrated at two locations up until the onset of the LERZ eruption in May: a lava lake at the summit in Halema 'uma 'u crater, active since 2008 (Patrick et al., 2019); and Puʻuʻōʻō in the middle East Rift Zone (MERZ) with activity lasting for 35 years (Patrick et al. 2020a).The shift to the LERZ in 2018 (Fig. 1a), which resulted in the loss of over 700 buildings to lava (Patrick et al. 2020b;Neal et al. 2019), marked a change in a long settled pattern of activity at Kīlauea.
A small short-lived fissure eruption at Puʻuʻōʻō, on the middle East Rift Zone (Fig. 1), on 30 April 2018 (Neal et al. 2019), was followed by earthquakes migrating eastward into the LERZ.This seismicity, marking the eastward migration of a shallow dike system, continued throughout the next 4 days, until the LERZ eruption began when magma reached the surface at 5:00 pm HST on 3 May in Leilani Estates, at the first of 24 fissures.On 4 May 2018, a M w 6.9 earthquake at a depth of 6 km (Neal et al. 2019) occurred on Kīlauea's south flank.Gansecki et al. (2019) described the three main phases of the eruption.During phase 1 (3-18 May), fissures 1-20 opened (Fig. 1a).These fissures were small and short-lived, opening in a semi-random distribution.Phase 2 (17-27 May) included the opening of fissures 21-24, alongside the continued Hawaiian fountaining at other eruptive fissures, producing lava flows to the southeast.Phase 3 (28 May-4 August) saw activity focus at fissure 8 producing a scoria cone, which, in 2021, was named Ahuʻailāʻau (Fig. 1;Hurley 2021;Deitterich et al., 2021).An increase of mass eruption rate during this phase resulted in a large damaging channelized lava flow travelling more than 13 km to the eastern coastline and covering an area of ~ 22 51 Page 4 of 11 km 2 (Fig. 1; Dietterich et al. 2021).Surface discharge of magma ended on 4 August and the eruption was officially declared over on 5 September 2018.More than 1 km 3 of lava was erupted over the course of the eruption (Fig. 1; Neal et al. 2019); ~ 92-96% of that was erupted from fissure 8 (Gansecki et al. 2019;Fig. 1a).
During phase 3, fissure 8 underwent an intensive and prolonged period of low Hawaiian fountaining (~ 80 m, Lyons et al. 2021) which formed the ~ 40 m high scoria cone (Fig. 1).In the first week of the episode, very-lowdensity clasts escaped the fountain and were carried more than 600 m downwind of the vent.This focused, intensive episode coincided with the arrival of hotter, more primitive magma from the MERZ (1145°C, Gansecki et al. 2019).The high magma discharge rate in late May constructed the tephra blanket and the lower portions of Ahuʻailāʻau cone (Fig. 1), where a lava pond rapidly formed from the collapsing fountaining.In June and July, fountaining jets played to a significantly lower heights with pulsating behaviour being observed both in the jets and in the adjacent fissure 8 lava flow (Patrick et al. 2018(Patrick et al. , 2019)).The detailed physical and geophysical observations, the sustained and high mass eruption rate and the highly destructive nature of the 2018 LERZ lava, makes the Ahuʻailāʻau deposits an extremely important study site for understanding the 2018 Hawaiian fountaining within the LERZ.

Methodology
Tephra thickness measurements (to the nearest cm) were carried out across the area of the continuous widespread fall deposit in February-March 2022, with all samples being obtained from within 650 m of the fissure 8 vent.The most proximal site was within 120 m of the eruptive vent.No bedding was present at the sample locations and hence no subunits have been defined.
Representative samples were collected, at 57 localities spread evenly across the study area (Figs. 4, 6), one at every location, in order to reconstruct TGSD analysis of the tephra blanket.Grain size samples were sieved down to − 3.2 Φ (in 10-mm intervals, see Supplementary Material S.2) in the field with the sieved fractions and the finer grained residue weighed with a field balance.A split of the residue (tephra > − 3.2 Φ, < 9.5 mm) was transported to the University of Hawaii at Manoa, and dried in a furnace at 110 °C for a minimum of 24 h.Samples were weighed pre-and post-drying to correct for moisture.The calculated moisture correction factor was then also applied to the mass of the corresponding coarser fractions of the sample sieved in the field.The dried samples then underwent further sieving, in 0.5 Φ steps down to + 2.0 Φ. Everything less than + 2.0 Φ accounted for less than < 3.5% for all except four samples and so was combined in one bin (> + 2.0 Φ).The four finer samples were sieved further down to + 2.5 Φ where the remainder > + 2.5 Φ then accounted for less than 3.5% of the total mass.Independently at each site, we collected the largest five clasts within 5 m of the grain size and measured (to the nearest mm) the longest, shortest, and intermediate axes for each clast and averaged.Number of sample sites were limited compared to other comparable studies (e.g.Mueller et al. 2018), however quantitatively still within acceptable limits as investigated by Klawonn et al. (2014a and2014b).
Density and bulk vesicularity analyses were carried out on two suites of 16-32-mm clasts (following  Houghton and Wilson (1989), and vesicularities calculated using a dense rock equivalent of 2800 kg m −3 .
Isopach maps of the widespread tephra blanket deposit were drawn by hand, with the most distal isopach being 10 cm.Thickness of deposits beyond this contour could not be constrained with confidence in the field due to reworking of the deposit through wind, weather, and human interference.The isopach areas were then calculated through Arc-GIS Pro.Deposit thinning and volume calculations were constrained using the TephraFits model (Biass et al. 2019b) which applies three empirical thinning relationships (Exponential, Power-Law and Weibull) to the available data.We specified a 7% uncertainty on the isopach areas when using the TephraFits model as this was approximated for a wellsampled medial deposit within Klawonn et al. (2014b).Overall uncertainty in the model fitting (using a 5 and 95% confidence interval calculated using an inbuilt Monte-Carlo approach) results in volume estimate errors of approximately ± 12% (Table 1).
Total grain size distribution of the widespread fall (i.e.excluding tephra forming the cone or incorporated into the lava flow) was estimated using the Voronoi tessellation Table 1 Volume estimates of the Fissure 8 tephra blanket deposit using three empirical thinning relationships: exponential (Fierstein and Nathenson 1992), power-law (Bonadonna and Houghton;2005), and Weibull (Bonadonna and Costa 2012;2013).Values calculated using TephraFits model (Biass et al. 2019a, b) Volume (m 3 ) R 2 Volume min (m 3 ) method (Bonadonna and Houghton 2005).Voronoi tessellation creates cells around every sample location, with each edge being a bisector of the line segment that would connect that sample location to a surrounding sample.This method ensures that the cell only includes grid points which are closer to those samples than any other.Once the cells have been created, each cell is assigned the deposit thickness, mass loading (calculated using a bulk density of 0.2 kg m −3 ), and grain size distribution (GSD) of the corresponding sample.The TGSD is obtained from the massloading and area-weighted average of all the cells within the deposit.The Voronoi TGSD results presented here were calculated using the MATLAB code TOTGS (Biass and Bonadonna 2014).The code requires a zero-isopachcontour, which was estimated based on extrapolation of the deposit; however, the calculation is not sensitive to the positioning of the zero-isopach-contour at the scale of the deposit within this study.
For each grain size dataset associated with a sampling site, a cumulative frequency line was drawn based on the raw data obtained through the sieving (see Supplementary Material S.2).Regular interval grain size values, the Inman coefficient ([Φ 84 -Φ 16 ]/2), and median diameter (Φ 50 ; Fig. 9) have been calculated by extracting the associated phi values (e.g.Φ 16 , Φ 50 , Φ 84 ) from the cumulative frequency curves for each individual sample (Fig. 7a).The modal grain size for each sample was assessed by looking at the phi values over which the cumulative frequency line had the greatest gradient.

General characteristics of the ejecta
The proximal products of the F8 fountain constitute the scoria cone of Ahuʻailāʻau (Fig. 1), which is dominated by large (i.e.> 64 mm) fluidal clasts (Fig. 2), with minor Pele's hair and tears, with fluidal textures being very common on all large clasts forming the surface of the cone, indicating the core of the bombs had not completely crossed the glasstransition, allowing them to continue to vesiculate, deform, and sometimes weld and/or flow after landing.However, this is not true about all the pyroclasts on the cone: some (scoria) landed in a solid state.A large range in vesicularity (36 to 95 volume %) was observed on the scoria cone, similar to that previously reported (i.e.41 to 96 volume %; Namiki et al. 2021).Namiki et al. (2021) proposed that the high vesicularities result from secondary vesiculation, as a consequence of rapid bubble nucleation and adiabatic gas expansion, subsequently causing further (i.e.secondary) fragmentation.
The widespread medial tephra blanket deposit is lapillidominated with a large proportion of reticulite and minor golden pumice (Fig. 2).Negligible abundances of larger fluidal clasts (Fig. 2) and Pele's tears and hair were also observed.Widely separated, decimetre-diameter, highly vesicular bombs were commonly observed but were often broken either from the impact upon landing, or from human influence post-eruption.Distal tephra, which formed a fringe surrounding the preserved medial fall, were, at the time of our mapping, very poorly preserved and have mostly been reworked or removed by residents during clean-up operations.

Vesicularity
The effect of the secondary fragmentation at the top of the highest F8 fountains, described by Namiki et al. (2021), is not only a significant reduction in clast size but also a major change in vesicularity and hence density of the clast population.The medial deposit is dominated by reticulite clasts (Fig. 2) with density 100-300 kgm −3 (Fig. 3), containing 91 to 97 volume % of c. 1-mm vesicles of very uniform size.A smaller subpopulation of clasts contains small vesicles together with scattered mm-cm bubbles (golden pumice, 82 to 91% vesicles, Fig. 2).Clasts from the cone are dominantly black scoria (Fig. 2) with a broad range of vesicularity from 42 to 79% vesicularity and density of 500-1750 kgm −3 (Fig. 3).Cone-forming clasts reflect only the primary magmatic fragmentation at the base of the fountain.Namiki et al. (2021) suggest that, in the medial deposit, the signature of primary fragmentation has been overprinted by a short but intense period of bubble nucleation accompanying secondary fragmentation at the top of the fountain.

Thickness and isopach map
The dispersal axis is 265° which aligns with the observed dominant easterly wind direction in Leilani Estates in May 2018.The thinning of the tephra blanket is approximately exponential.The thickness of the deposit decreases by a factor of two approximately every 60 m when on-axis and within 300 m of the vent.The most distal deposits have a thinning distance of approximately 90 m, where the deposit is about 10 cm thick.
The isopach map (Fig. 4) of the tephra blanket is based on field thickness data collected by the authors (see Supplementary Material S.2).Thickness considerations meant that Ahuʻailāʻau cone and the ejecta that formed the rheomorphic lava flow were not included in the isopach map or volume estimates.The closest data point was ~ 120 m from the erupting vent (Fig. 3).Five isopach contours have been closed (Fig. 4), with the smallest isopach (10 cm) covering an area of 0.22 ± 0.009 km 2 .This was the smallest isopach recorded due to the poor preservation of the thin deposits falling in the adjacent residential environments.The four other isopach contours are very well constrained (Fig. 4; Supplementary Material S.1) with the areas of each shown in Supplementary Material S.2.A few thickness measurements were not included in the isopach map as they appeared to be anomalies.These anomalies can be explained due to two reasons: (1) proximity to cleared access roads and thus likely to have been affected by human interference; (2) shadow effects (i.e.atmospheric turbulence) as a result of some subsidiary cones (Fig. 6).

Volume estimates
Volume estimates of the tephra blanket (not including Ahuʻailāʻau and the lava flow) have been calculated based on the isopach areas using the TephraFits model (Biass et al. 2019b) and are shown in Table 1 and Fig. 5.The TephraFits model applied three empirical thinning relationships (exponential, power-law and Weibull) to the isopach contour areas.
All three relationships offer excellent fits to the data (R 2 = 0.99) despite the lack of distal data.The lack of distal constraints was a result of the modification of the deposit by wind/weather erosion and most commonly human interference due to the deposit falling on a housing estate.The total volume of the medial deposit is estimated to be 1.57× 10 4 ± 0.19 m 3 (exponential), 2.13 × 10 4 ± 0.22 m 3 (Weibull), and the best fit thinning relationship (power-law) returned a volume of 1.99 × 10 4 ± 0.23 m 3 (Fig. 5b).This is just 0.02% of the total volume erupted from the main episode of F8 in 2018 (Neal et al. 2019); the remainder is contained within the proximal cone and the rheomorphic lava flow.
Particles finer than 1.5 Φ make up a maximum of 3.5 wt% of the samples.The grain size range of the analysed samples is − 6 to 2.5 Φ (64-0.2mm).Bombs ranging in size from − 8.6 to − 6 Φ (410 to 64 mm) could be found on the surface of the tephra deposit at (within 5-m radius) every sample location.The bombs were often broken; as a result, they were never observed deeper in the deposit.When mapping the sizes of the largest clasts across the study area, the distribution does not show a clear pattern and hence no isopleth lines have been plotted (Fig. 6).Grain size distribution (GSD) of single samples is consistent across the whole study area showing minimal fining with distance from the vent (Fig. 7a).Lapilli (− 5.5 < Φ < − 1) dominated the GSD contributing 80.9-96.3wt% to the deposit volume in all samples.The mode of all the measured samples was within the range of − 4 to − 3.5 Φ with 60% of the samples having a modal grain size of − 3.5 Φ and the remaining 40% having a mode grain size of − 4 Φ.The amount of tephra estimated from the Voronoi tessellation TGSD (cells shown in Fig. 7c) was 91.8 wt% lapilli and the mode was − 3.5 Φ.A few anomalously coarse grain size distributions can be observed both close to (160-170 m) and farther from vent (> 500 m).The anomalously coarse samples away from the vent are frequently in locations that have been obstructed by the minor scoria cones southeast of Ahuʻailāʻau indicating complex aerodynamics and reentrainment of large clasts close to the ground, occurred, disrupting the typical tephra settling pattern.Rare finer samples can also be observed (Fig. 7a), which may reflect (1) contamination by late-stage ash through re-dispersal by the wind or (2) human activity, crushing clasts producing a high proportion of fine tephra.

Tephra production and re-entrainment
The anomalous thickness measurements (see the "Thickness and isopach map" section) and sample sizes (see the "Grain size distribution" section) at fissure 8 could be explained through two main reasons: (1) the four/five subsidiary cones to the southwest (Fig. 6) created a shadow zone in which near surface wind gusts formed an local atmospheric circulation system, which re-entrained large clasts that were carried beyond the subsidiary cones (Fig. 8).Re-entrainment would have been facilitated by a secondary vesiculation event as described by Namiki et al. (2021).Secondary vesiculation produced the dominant population reticulite/golden pumice clasts with extremely low density, which in turn enabled their re-lifting by upwelling air parcels (i.e.local atmospheric circulatory systems), much like the movement of a feather (Fig. 8).( 2) The duration of time since the tephra has been deposited opened up the possibility of fragmentation and degradation of the deposit through human activity (i.e.people walking over the deposit, and clearing tephra off their land).We propose that it was a combination of these mechanisms that resulted in the high modal grain size and anomalously coarse particles of the deposit observed today.
It should be noted that whilst it had only been 4 years since the F8 eruption when carrying out the field campaigns, human interference (i.e.clearing the deposit from properties/roads and walking/driving over deposit) was evident.In this study, it was clear where this had occurred and hence these areas were avoided, thereby minimising the impact of these interferences on the calculations in this study.This human interaction likely increases over time as people regain the land, and therefore, in future studies

Comparisons with other lava-fountaining eruptions
There are no well-studied medial deposits of other lava-fountaining eruptions that match the grain size characteristics of the F8 deposits closely.In general, the 20 well-studied deposits collated by Costa et al. (2016) are for other more powerful eruption styles and are typified by two overlapping grain size populations.This bimodality is inferred to reflect contrasting transport modes but not necessarily different fragmentation histories, as inferred by Namiki et al. (2021) for the F8 deposits.Of the 20 deposits used by Costa et al. (2016), only one (Etna 2007) is of moderate Hawaiian intensity (Fig. 9).The most intensively sampled Hawaiian medial fall is that of Kīlauea Iki in 1959 (Richter et al. 1970;Klawonn et al. 2014aKlawonn et al. , 2014b;;Mueller et al. 2018).The mode of its TGSD is − 2 phi, one phi unit finer than the F8 medial deposit.This is predictable given that the 1959 fall was the product of generally higher lava fountaining and was constrained out to 4 km from vent (Mueller et al. 2019).The widespread deposit mostly contains a mixture of golden pumice and fluidal clasts, with wider ranges of vesicle sizes than present in reticulite, which was interpreted as the products of a single primary vesiculation event over a prolonged time interval extending past the time of fragmentation (Stovall et al. 2011;2012).A much smaller amount of true reticulite was erupted (Richter et al. 1970) but mostly blown by the strong winds beyond the limits of the continuous tephra sheet.At the peaks of the highest fountaining episodes (3, 15, 16), dense scoria clasts were also transported beyond the 1959 cone, to distances up to 500 m from the vent.
In general, the more powerful deposits that Costa et al. (2016) collated are typified by two overlapping grain size populations inferred to reflect contrasting ballistic and convective transport.Ballistic clasts do not extend into the medial 2018 LERZ products, and the distinction between the our data and that from the other lava-fountaining eruptions cited by Costa et al. is a function of our inability to sample the coarse-grained, but intensely welded 2018 cone.

Conclusions
Very high clast vesicularities during the final phase of lava fountaining from F8 in 2018 led to distribution of the pyroclasts as a widespread fall beyond the proximal scoria cone.This subpopulation of exceptionally light pyroclasts was able to escape from the top of the fountain with the decoupled gas phase whilst the numerically dominant denser clasts fell from the fountain to form the cone and fed the associated lava flow.The very light reticulite/golden pumice clasts were strongly affected by fluctuations and eddies in the near surface wind field, being re-entrained into plumes especially in the lee of subsidiary vents to the southwest of the cone (Fig. 8).The lack of a systematic fall-off in maximum clasts size with distance from vent which is typical for pyroclastic fall deposits (Sigurdsson et al. 2015) can be explained by this re-entrainment process.

Fig. 1 a
Fig. 1 a Lava flows from the 2018 lower East Rift Zone eruption.The red flow is the main flow from fissure 8, and the orange are all other flows associated with the 2018 eruption.The star shows the location of the Ahuʻailāʻau cone.The fissures are shown as blue lines and the area of this study is outlined by the dotted box.GIS data was accessed through the USGS science data catalog: Fissure locations and lava flow areas (Zoeller et al. 2020) Lidar Digital Elevation

Fig. 2
Fig. 2 Photos of typical pyroclasts obtained from the cone-forming (a, b) and medial (c, d) F8 tephra deposit.a Fluidal pyroclast ~ 10 cm in size.b Scoria ~ 6 cm in size.c Golden pumice ~ 7 cm in size, d reticulite ~ 6 cm in size

Fig. 3 Fig. 4
Fig. 3 Distribution of density within the pyroclasts from the medial deposit (light blue) and the Ahuʻailāʻau cone (dark blue).Mean for each deposit is shown by the dashed and dot-dashed lines respectively

Fig. 5 aFig. 6 51
Fig. 5 a Log thickness vs square root of area plot showing the thinning of the deposit.Crosses represent the actual observed thickness points and the three fits, i.e. exponential, power-law, and Weibull are displayed by blue, yellow and green lines respectively.The 5 h and 95th percentile confidence interval are shown by dotted lines in the corresponding colour.b Box and whisker plots for volume estimates of each of the empirical thinning laws.Colours correspond to the style of fit as in a.The box shows the 1st and 3rd quartiles.The ends of the whiskers show the 5th and 95th percentile

Fig. 7 a
Fig. 7 a Cumulative frequency plot, showing the grain size distribution of 55 samples colour-coded with distance from the vent.Yellow being medial and blue being close to vent.Black dashed line shows the results from Voronoi TGSD (labelled TOTGS).b Values of the 16th (yellow circles), 50th (green diamonds), and 84th (blue squares) percentiles for all 55 samples illustrating how the percentiles used as statistic parameters for sorting vary with distance from the eruptive vent.c Map showing the Voronoi cells, applied to the tephra blanket of F8, used to calculate the TGSD.Each cell represents a Voronoi cell built around each sample point (blue dots).The outermost line represents the estimated line of zero mass

Fig. 8 Fig. 9
Fig. 8 Schematic overview of the Fissure 8 vent Hawaiian fountaining up to 80 m high.Secondary vesiculation within the ejecta results in further fragmentation (referred to as secondary fragmentation).The ejecta can be seen to be dispersed with the low-density pyroclasts (reticulite/golden pumice) being re-entrained by the local upwelling air parcels resulting in an unusual distribution of medial clasts.Adapted fromMueller et al. (2018)