The length of channelized lava flows: Insight from the 1859 eruption of Mauna Loa Volcano, Hawai‘i

Journal of Volcanology and Geothermal Research 183 (2009) 139–156

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j vo l g e o r e s

The length of channelized lava flows: Insight from the 1859 eruption of Mauna Loa Volcano, Hawai‘i Jenny M. Riker a,⁎,1, Katharine V. Cashman a, James P. Kauahikaua b, Charlene M. Montierth c a b c

Department of Geological Sciences, University of Oregon, Eugene, OR 97403, USA U.S.G.S. Hawaiian Volcano Observatory, Hawai‘i Volcanoes National Park, HI 96718, USA Clark College, Vancouver, WA 98663, USA

a r t i c l e

i n f o

Article history: Received 6 October 2008 Accepted 2 March 2009 Available online 16 March 2009 Keywords: lava flow Mauna Loa crystallization channels geothermometry

a b s t r a c t The 1859 eruption of Mauna Loa Volcano, Hawai'i, produced paired 'a'ā and pāhoehoe flows of exceptional length (51 km). The 'a'ā flow field is distinguished by a long (N 36 km) and well-defined pāhoehoe-lined channel, indicating that channelized lava remained fluid to great distances from the vent. The 1859 eruption was further unusual in initiating at a radial vent on the volcano's northwest flank, instead of along the welldefined rift zone that has been the source of most historic activity. As such, it presents an opportunity both to examine controls on the emplacement of long lava channels and to assess hazards posed by future flank eruptions of Mauna Loa. Here we combine evidence from historical chronicles with analysis of bulk compositions, glass geothermometry, and microlite textures of samples collected along the 1859 lava flows to constrain eruption and flow emplacement conditions. The bulk compositions of samples from the 'a'ā and pāhoehoe flow fields are bimodally distributed and indicate tapping of two discrete magma bodies during eruption. Samples from the pāhoehoe flow field have bulk compositions similar to those of historicallyerupted lavas (b 8 wt.% MgO); lava that fed the 'a'ā channel is more primitive (N 8 wt.% MgO), nearly aphyric, and was erupted at high temperatures (1194–1216 °C). We suggest that the physical properties of proximal channel-fed lava (i.e., high-temperature, low crystallinity, and low bulk viscosity) promoted both rapid flow advance and development of long pāhoehoe-lined channels. Critical for the latter was the large temperature decrease (~50 °C) required to reach the point at which plagioclase and pyroxene started to crystallize; the importance of phase constraints are emphasized by our difficulty in replicating patterns of cooling and crystallization recorded by high-temperature field samples using common models of flow emplacement. Placement of the 1859 eruption within the context of historic activity at Mauna Loa suggests that the formation of radial vents and eruptions of high-temperature magma may not only be linked, but may also be a consequence of periods of high magma supply (e.g., 1843–1877). Flank eruptions could therefore warrant special consideration in models and hazard mitigation efforts. © 2009 Elsevier B.V. All rights reserved.

1. Introduction On January 23, 1859, an eruption of Mauna Loa Volcano, Hawai'i, initiated at several vents high on the mountain's northwest flank (Fig. 1). During the first 8 days of the eruption, a channelized 'a'ā flow traveled 51 km to the western coast of the Island of Hawai'i, producing not only the longest lava flow in the volcano's historic record, but also a long and well-developed pāhoehoe-lined channel (36 km). A parallel tube-fed pāhoehoe flow formed over the following 10 months, making the 1859 eruption one of several historic eruptions to generate paired 'a'ā and pāhoehoe flows (e.g., Rowland and Walker, 1990). This eruption was unusual in many respects. First, its flows initiated at a radial flank vent, instead of along the two rift zones that have been the

⁎ Corresponding author. Tel.: +44 29 2087 4573; fax: +44 29 2087 4326. E-mail address: [email protected] (J.M. Riker). 1 Present address: Department of Earth Sciences, University of Bristol, Wills Memorial Building, Queen's Road, Bristol, BS8 1RJ, United Kingdom. 0377-0273/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jvolgeores.2009.03.002

source of most other historic eruptions (Fig. 1). Second, it produced magma of heterogeneous composition (Rhodes, 1983); this contrasts with the remarkable compositional homogeneity that has typified Mauna Loa's historic and prehistoric eruptions (Powers, 1955; Wright, 1971; Rhodes, 1983; Rhodes, 1995; Rhodes and Hart, 1995). Third, it produced anomalously long channelized 'a'ā and tube-fed pāhoehoe flows. Finally, the 1859 eruption occurred during an uncommonly active four-decade period at Mauna Loa (1843–1877; Lockwood and Lipman, 1987). Here we suggest that these seemingly disparate characteristics (flow location, composition, length, and timing) may be related. Moreover, we demonstrate ways in which an improved understanding of this eruption can aid assessment of lava flow hazards related to rare radial vent eruptions on the flanks of Mauna Loa Volcano. The unique hazards posed by flank eruptions have been studied at Mount Etna, Italy (Acocella and Neri, 2003), where they may initiate along rift zones or at isolated radial vents and tend to have both higher eruption rates and larger total volumes than summit eruptions. Flows

140

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

Fig. 1. Simplified map of Mauna Loa Volcano (shaded region) on the Island of Hawai'i, showing lava flows produced since 1843 (‘historic eruptions’). Dashed lines show approximate location of summit caldera and rift zones. Labeled eruptions (dark grey) are discussed in this paper. Lava flows from the 1859 eruption are distinguished by their exceptional length (51 km) and off-axis vent location. HUA = Hualālai, MK = Mauna Kea, and KIL = Kīlauea.

with high eruption rates tend to travel farther than those with lower ones (Walker, 1973), while flows of large volume may cover more area than smaller flows. Combined with the effect of flank vents being displaced from the summit (toward population centers), it is clear that flank events at Mount Etna pose greater risks than typical summit eruptions. Is the same true for Hawaiian volcanoes? Unlike Mount Etna, the overwhelming majority of Mauna Loa's eruptive vents form along two well-established rift zones. The rarity of historic eruptions on the northwest flank (hereafter termed ‘flank,’ ‘radial,’ or ‘off-axis’ eruptions) prevents an equivalent analysis of the hazards they pose. However, displacement of radial vents from the summit and upper rift zones toward populated coastlines clearly increases hazard, both because of increased probability that flows will invade populated areas, and because proximal flow advance rates are generally higher than those in distal regions (e.g., Kauahikaua et al., 2003; Soule et al., 2004). Additionally, the exceptional length of the 1859 'a'ā flow, along with its well-formed, pāhoehoe-lined channel, give evidence of sustained lava fluidity, raising questions about the nature of flank vent eruptions. This is of particular concern as it relates to channelized lava flows, which advance rapidly (up to 4 m/s; Kauahikaua et al., 2003) and may transport lava at very high speeds (up to 15 m/s; Lipman and Banks, 1987). Of the 29 flows produced by historic eruptions of Mauna Loa (Fig. 1), 23 were channel-fed (Rowland et al., 2005). Understanding controls on the emplacement of channelized lava flows is therefore critical to effective hazard assessment and response. Over the past few decades, studies have established key factors governing the lengths and advance rates of lava channels including effusion rate (e.g., Walker, 1973; Pinkerton and Wilson, 1994; Kauahikaua et al., 2003), topography (e.g., Kilburn, 1990; Kilburn, 1993; Pinkerton and Wilson, 1994), and cooling and crystallization (e.g., Crisp et al., 1994; Cashman et al., 1999; Soule et al., 2004).

Effusion rate is generally considered to exert a primary control on final flow length (Walker, 1973; Pinkerton and Wilson, 1994; Kauahikaua et al., 2003), although effusion rate alone is a poor predictor of the lengths of lava channels at Mauna Loa and Kīlauea (Fig. 2), which may be better correlated with flow volume or eruption duration (Malin, 1980; Pinkerton and Wilson, 1994). Moreover, none of these

Fig. 2. Length versus effusion rate for channel-fed Hawaiian lava flows (modified from Pinkerton and Wilson 1994). All data except those for the 1859 and 1984 eruptions are from Malin (1980). For consistency with Malin (1980), effusion rates are ‘actual effusion rates’ (flow volume divided by the length of time the flow was actively fed). Dashed lines are flow length limits empirically defined by Walker (1973) based on a global data set. The effusion rate estimate for the 1859 channel is discussed in the text. Effusion rates for 1984 flow lobes 1 and 1A were estimated from the cumulative flow volumes and durations given by Lipman and Banks (1987). Flows less than 45 h in duration are assumed to be supply-limited and are excluded (e.g., Pinkerton and Wilson, 1994).

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

explanations adequately account for the tendency of Hawaiian lava channels to reach maximum lengths of ~25 km over a wide range of effusion rates (Fig. 2). While partly a consequence of the island's size (see ‘ocean-limited’ flows in Fig. 2), this observation also holds true for many flows that did not reach the coastline. We use samples collected along the lengths of the 1859 lava flows to explore other parameters that might influence lava flow length, including composition, eruption temperature, conditions of cooling and crystallization, and topographic effects. Our primary focus is the 1859 'a'ā channel, which plots at the upper bounds of Walker's limits and exceeds by a factor of two the lengths of other channel-fed Hawaiian lava flows (Fig. 2). Observational and petrologic data from the well-documented 1984 eruption of Mauna Loa, which generated a channel of more typical length (27 km), provide a basis for comparison and calibration of this study. While our goal is to shed light on lava flow dynamics, particularly as they apply to hazard assessment, we also explore the relationship between the 1859 eruption's many unusual characteristics (i.e., length, volume, effusion rate, and composition) and the equally unusual period of activity in which it occurred. 2. The 1859 eruption of Mauna Loa Volcano The chronology of the 1859 eruption is well-documented, thanks to the careful observations of surveyor W. D. Alexander (1859), writers T. Coan and L. Lyons (Dana, 1859), and other contemporary observers (Haskell, 1859a,b; Haskell, 1860, reprinted in Barnard, 1990). Numerous newspaper accounts also provide insight into the sequence of eruptive events (Kaakua, 1859; Lyman, 1859; Lyons, 1859; NaKapae, 1859). The eruption began in the evening of January 23, apparently near the summit, as fountaining was visible from both sides of the island (Waimea and Hilo; Dana, 1859; Haskell, 1859b; Lyman, 1859). Early activity focused at a series of radial vents high on Mauna Loa's northwest flank (~ 3400 m, here denoted the ‘upper vents’; Fig. 3). A second lava breakout occurred immediately down slope several hours later (~ 2900 m, ‘lower vents’; Fig. 3). Historic accounts suggest that activity at higher elevations waned after the first day of the eruption, although diversion of flows from the upper vents around cinder cones at the lower vents suggests that lava continued to flow from the upper vents for at least several days. All subsequent activity appears to have initiated at the lower vents (e.g., Lyman, 1859). The origin of a third fissure system to the north (~ 2700 m, ‘north vents’; Fig. 3) is unclear. However, L. Lyons (Dana, 1859) noted that “two streams of fire were issuing from two different sources, and flowing, apparently, in two different directions” at the outbreak of the eruption on January 23. Given their relative proximity to Hilo, these vents may in fact be the source of ‘summit’ activity reported by Haskell (1859b) and other observers on the west side of the island. As later accounts make no mention of the north vents, we suggest that they were also active only at the commencement of the eruption. The first phase of the 1859 eruption produced a channel-fed 'a'ā flow that reached the ocean on the evening of January 31, after just eight days of travel. The flow front had crossed the coastal road (current Hwy 19, less than 1 km from the coast) that morning (Dana, 1859). The 'a'ā flow field is distinguished by a prominent channel (hereafter termed the ‘primary channel’) that is well-defined along most of its length (Fig. 3). The ubiquitous presence of pāhoehoesurfaced overflows suggests that channelized lava remained fluid at great distances from the vent, consistent with contemporary observations (Dana, 1859; Haskell, 1859a). Historic accounts provide useful descriptions of channel development. Visitors to the lower vents on February 9 (day 18 of the eruption) describe an open channel “about half a mile below the lower of the two craters” emitting lava “at a white heat and apparently as liquid as water” (Haskell, 1859a). Both Haskell and Alexander note that the channel was well-confined, with minimum dimensions

141

of 6–15 m width and 3–5 m depth for the first few miles (Alexander, 1859), after which “it gradually changed into a net-work of streams” and the color of the lava surface changed to a “deep blood-red” (Haskell, 1859a). Similarly, Lyman (1859) reports that “instead of running in one large stream, [the channel] parts and divides into a great number – perhaps as many as fifty – spreading out over a tract of five or six miles in width.” He estimates flow velocities of 4–10 miles/h (2.8–4.4 m/s) within 10 km of the vents, rates that decreased in the saddle between Mauna Loa and Hualālai Volcanoes as the primary channel developed a “very tortuous” path. Flow advance rates in the saddle were estimated at 2–3 miles/h (~ 1 m/s). By mid-February (no earlier than February 9, day 18; Haskell, 1859b), fountaining at the lower vents had either ceased or become intermittent, and several observers commented that surface flows had slowed or ceased by the end of the first (Lyman, 1859) or second (Kahookaumaha, 1859) week of February. Observers visiting the source after day 27 (February 18) describe a tube system originating at the lower vents (Haskell, 1859b; Sleeper, 1859). The duration of the channel-fed ('a'ā) phase can therefore be constrained to between 18 and 27 days. The second phase of the eruption produced a tube-fed pāhoehoe flow that also reached the ocean. Although the exact date of ocean entry is unknown, the flow front was clearly still in the saddle region in late February (Sleeper, 1859), where it buried earlier-formed channels; it probably reached the coast in late June (NaKapae, 1859; Gower, 1886) and was active through November of 1859 (Haskell, 1860), for a total eruption duration of about 10 months. During this time it covered the large Kiholo fish pond (Kaakua, 1859; Coan, 1860) and numerous archaeological sites such as the Hōlua slide (Ching, 1971) and the Heiau (temple) of Lono'akai in north Kona (Stokes, 1991). The flow field also inspired an early description of pillow basalt formation (‘spheroids,’ Green, 1887). Taken together, the activity of 1859 produced one of Mauna Loa's most voluminous historic flows, with an estimated subaerial volume of 383 × 106 m3, of which 270 × 106 m3 is 'a'ā and 113 × 106 m3 is pāhoehoe (Rowland and Walker, 1990). Lockwood and Lipman (1987) estimate an additional submarine volume of ~ 95 × 106 m3. Scuba diving reveals that the central submarine portion of the southern (tube-fed) 1859 flow is made up of coherent pillow lava, including well-formed shelly pillows (J. Moore, pers. comm., Jan. 2009). However, no submarine observations are available for the channel-fed flow, and it is not possible to assign accurate submarine volumes to either phase of the eruption. Assuming that 'a'ā production ceased between day 18 and day 27, average eruption rates were ~116–235 m3/s during the 'a'ā phase of the eruption and ~5–9 m3/s during the pāhoehoe phase (where maximum estimates for each phase include the entire submarine volume and minima exclude the submarine volume). Our estimates for the 'a'ā phase of activity differ from those previously reported (Rowland and Walker, 1990), as we have included the submarine flow volume in our calculations and our interpretation of historic accounts permits a somewhat longer duration of channel-fed activity. The constraint of 8 days travel time to the ocean yields an average 'a'ā flow advance rate of 267 m/h, or 0.07 m/s. It also allows us to estimate an eruption rate of ~391 m3/s during the period of active channel lengthening. 3. Mapping and sampling The surface distribution of 'a'ā, pāhoehoe, and vent facies in the 1859 flow fields has previously been mapped by Rowland and Walker (1990). We present an updated map in Fig. 3, in which we have used digital orthophotos to map the distribution of channels within 'a'ā flows, as well as ‘skylights’ marking the trace of lava tube networks that fed the pāhoehoe phase of the eruption (Fig. 3a). Field work focused on a detailed investigation of vent regions, with the specific goal of resolving temporal and compositional relationships between

142

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

Fig. 3. Map of the 1859 eruption of Mauna Loa. (a) Map of the 1859 lava flows, showing sample locations, channels, skylights, and the surface distribution of pāhoehoe and 'a'ā. Inset aerial photographs show the diversion of upper vent 'a'ā flows around cones at the lower vents (upper inset) and a typical channel segment lined with pāhoehoe overflows (lower inset). (b) Detail of the proximal flow field. The three vent areas distinguished in this study are labeled. (c) Detail of the distal flow field.

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

the upper, lower and north vents (Fig. 3b) and the channel- and tube-fed flows. We also sampled the saddle between Mauna Kea and Hualālai, where late-phase pāhoehoe covers the primary channel and the two

143

flow fields are complexly branched and interfingered. Finally, we sampled at regular intervals along the lengths of the 'a'ā and pāhoehoe flows to check the results of a preliminary study of distal samples

Table 1 Thermal and textural data for lava samples from the 1859 eruption. Sample 901-6a 901-3a M-4 M-1b M-2b 901-2a M-10 901-6p 901-6ap 21-1 21-4f 901-5ap 901-4ap M-3a M-3b M-2a1 M-2a2 M-2c M-2d 21-2 901-5a 21-3 901-5p 901-3p M-6a M-6b M-8 M-9 901-2p 901-1p M-11c M-96a M-96b M-12a M-12b M-13 901-8e 28-5b 28-5c 901-8sp 28-2d 28-1 901-8de 901-8ce 901-8b2e 901-8b1e 28-5a 28-2c 901-7sp 28-3bf 28-4 28-2a 901-7p 28-3a 28-9b 28-8 28-6a 28-9c 28-9a 28-6b2 28-6b1 28-7 a b c d e f

Typea 'a'ā 'a'ā 'a'ā 'a'ā 'a'ā 'a'ā 'a'ā phh aphh aphh aphh aphh aphh aphh aphh aphh aphh aphh aphh 'a'ā 'a'ā phh phh phh phh phh phh phh phh phh phh phh phh phh phh phh sp sp sp sp sp sp sp 'a'ā aphh aphh phh phh sp sp aphh phh phh phh sp sp sp sp phh phh phh phh

Source Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Channel Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Tube Lower vents Lower vents Lower vents Lower vents Lower vents Lower vents Lower vents Lower vents Lower vents Lower vents Lower vents Lower vents Upper vents Upper vents Upper vents Upper vents Upper vents Upper vents North vents North vents North vents North vents North vents North vents North vents North vents

Distanceb

MgObulk

MgOglass

Tc

(km)

(wt.%)

(wt.%)

(°C)

10 32 34 35 37 40 45 10 10 21 21 22 28 34 34 35 35 37 37 21 22 21 22 32 34 34 35 37 40 44 45 47 47 49 49 49 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1 b1

9.29 8.72 n.d. n.d. 9.43 9.42 9.49 9.45 9.29 9.28 9.75 9.48 9.70 9.42 9.23 n.d. n.d. 10.06 n.d. 7.67 7.48 7.76 7.67 7.74 n.d. n.d. n.d. 7.84 7.83 7.71 n.d. 7.71 7.86 7.18 7.78 n.d. 9.82 10.05 n.d. 9.75 n.d. 9.23 9.69 9.72 9.71 9.69 n.d. n.d. 7.64 7.22 7.41 n.d. 7.58 n.d. 7.81 n.d. n.d. 7.81 n.d. n.d. n.d. n.d.

n.d. n.d. 5.66 5.54 5.71 n.d. 5.82 6.27 6.45 6.25 6.18 6.16 6.16 5.84 5.92 5.97 5.92 5.87 5.90 n.d. n.d. 6.35 5.99 5.94 5.99 5.98 6.09 6.18 5.95 6.10 6.20 6.16 n.d. 6.18 6.17 6.36 8.87 8.75 8.59 8.64 8.21 7.89 6.25 6.21 6.58 6.24 7.53 6.64 6.58 5.96 6.68 6.67 6.29 6.20 6.79 6.44 6.21 5.35 6.52 6.12 5.87 5.55

n.d. n.d. 1142 1139 1143 n.d. 1146 1156 1160 1156 1154 1154 1154 1146 1148 1149 1148 1147 1148 n.d. n.d. 1158 1150 1149 1150 1150 1152 1154 1149 1152 1155 1154 n.d. 1154 1154 1158 1216 1213 1210 1211 1201 1194 1156 1155 1163 1156 1185 1165 1163 1149 1166 1165 1157 1155 1168 1160 1155 1135 1162 1153 1147 1140

'a'ā ('a'ā), pāhoehoe (phh), aphh (channel overflow), sp (spatter). Distance from the lower vents. Calculated from glass MgO (wt.%) using the geothermometer of Montierth et al. (1995). Normalized to exclude vesicles and phenocrysts. Sample poorly quenched. Heterogeneous glass composition (1σ relative standard deviation of MgO measurements N2.5%).

ϕplagd

ϕpxnd

ϕmd

Na

plag

(mm− 2) n.d. n.d. 0.14 0.20 0.15 n.d. 0.12 0.07 0.05 0.05 n.d. 0.05 0.08 0.14 0.12 0.11 0.10 0.12 0.12 n.d. n.d. 0.03 0.12 0.12 0.13 0.11 0.15 0.11 0.12 n.d. 0.10 0.13 n.d. 0.13 0.11 0.11 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 0.00 0.01 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. 0.13 0.15 0.14 n.d. 0.12 0.06 0.03 0.06 n.d. 0.03 0.06 0.13 0.11 0.09 0.11 0.11 0.09 n.d. n.d. 0.01 0.04 0.09 0.07 0.08 0.08 0.06 0.09 n.d. 0.06 0.08 n.d. 0.07 0.07 0.06 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 0.07 0.19 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. 0.27 0.35 0.29 n.d. 0.24 0.13 0.08 0.11 n.d. 0.08 0.13 0.28 0.23 0.20 0.20 0.24 0.21 n.d. n.d. 0.04 0.16 0.20 0.20 0.19 0.23 0.18 0.20 n.d. 0.17 0.21 n.d. 0.20 0.18 0.17 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 0.07 0.20 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. 2300 2400 2500 n.d. 3600 419 517 1224 n.d. 1428 1990 2600 2500 2400 2600 2500 2400 n.d. n.d. 69 280 132 170 170 220 300 152 n.d. 220 180 n.d. 250 230 160 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 10 159 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

144

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

(Montierth, 1999), which suggested that the two flow fields had different bulk compositions. Vent relations are complex, with scoria and tephra produced by lava fountains intermingled with vent overflows and the proximal facies of the main channel. To relate distal samples to individual vents, we were careful to sample the full range of tephra and lava types from each vent location (upper, lower, and north; Fig. 3b). Our samples thus include as much of the proximal stratigraphic record as possible, with spatter and scoria from cones and tephra aprons in addition to glassy rinds of proximal flow lobes (typically of pāhoehoe morphology). Samples collected along the lengths of the two flow fields (Fig. 3) span the range of surface morphologies present at any given location: pāhoehoe flow crust, pāhoehoe overflows from the primary 'a'ā channel, and rough 'a'ā clinkers. Quenched glass is well preserved in the outer rinds of pāhoehoe-surfaced samples but could also be found at the outermost surfaces (often only the outer 1 or 2 mm) of some 'a'ā clinkers. We assume that these samples approximate the state of lava during flow, as tested by comparative textural studies of syn- and post-emplacement samples from recent Hawaiian eruptions (Crisp et al., 1994; Jurado-Chichay and Rowland, 1995; Cashman et al., 1999; Montierth, 1999). In particular, these studies show that pāhoehoesurfaced overflows on channel margins preserve the crystal textures of lava in the fluid channel core. Although temporal relationships among channel overflow samples are unknown, exposed surfaces probably represent a late stage of high lava flux, thus providing information on flow through the mature channel. 4. Analytical methods 4.1. Bulk compositions Bulk compositions of selected samples (Appendix 1, supplementary material) were measured by both XRF at the USGS laboratory in Denver, CO (C. Thornber, unpublished data), and by electron microprobe analysis of experimentally melted charges (this study). For the latter, a small portion of each sample was ground to a fine powder, mixed with polyvinyl alcohol to produce a slurry, and melted in a Deltech vertical quench furnace at 1335 °C for 4–5 h. A downward flow of mixed H2 and CO2 gas kept the furnace at either quartz–fayalite– magnetite (QFM) or nickel–nickel oxide (NNO) buffer conditions (as noted in Appendix 1). Water-quenched charges were analyzed for major and minor elements on a four-spectrometer Cameca SX-50 microprobe at the University of Oregon operated at a 15 keV accelerating voltage and a 20 nA beam current. We used a diffuse beam (10 µm diameter) to minimize alkali volatilization. The volcanic glass standard VG-2 was analyzed as an unknown at the beginning of each run to assess analytical accuracy, and data were reduced using the matrix correction of Armstrong (1988). Reported compositions are averages of 10–15 analyses per charge. 4.2. Glass compositions and geothermometry Compositions of glassy flow rinds (Appendix 2, supplementary material) were also obtained by electron microprobe analysis of 10–15 spots within the best-quenched portion of each sample, determined by visual inspection of microlite abundance. Quenching temperatures of all samples containing sufficient glass for microanalysis, as dictated by beam diameter, were then calculated using a composition-based geothermometer empirically calibrated for Mauna Loa lavas (Montierth et al., 1995). This geothermometer relates temperature to MgO (wt.%) in melt or glass as
o  T C = 23:0 × ðwt:k MgOÞ + 1012;

ð1Þ

recording the linear incorporation of MgO into olivine or augite + pigeonite across the crystallization interval. Propagated analytical

uncertainties yield an estimated geothermometer error of ±10 °C (Montierth et al., 1995), and comparative studies show that calculated temperatures typically fall within 10 °C of those measured in the field (Helz et al., 1995; Montierth et al., 1995). Larger systematic errors may be caused by poor quenching, resulting in low-temperature estimates (e.g., Helz et al., 2003), or by disequilibrium crystallization (induced by degassing, undercooling, or kinetic delays), resulting in hightemperature estimates (e.g., Helz et al., 1995; Montierth et al., 1995). Samples with clear textural indicators of poor quenching, such as pyroxene dendrites or pyroxene overgrowths on plagioclase, are indicated in Table 1, and their temperatures should be taken as minimum values. Heterogeneous glass compositions are typical of variably-quenched samples and will also yield minimum values; however, most analyzed glasses were compositionally uniform, with a relative standard deviation of spot analyses of b2.5% (1σ) (corresponding to b3.5 °C absolute uncertainty; two exceptions are noted in Table 1). Evidence of disequilibrium crystallization is discussed as appropriate below. 4.3. Textural analysis Backscattered electron (BSE) images of polished thin sections were used for qualitative and quantitative characterization of groundmass textures. Images were captured using a JEOL 6300V scanning electron microscope (SEM) at the University of Oregon operated at a 10 keV accelerating voltage, a 5 nA beam current, and a 15 mm working distance. Image acquisition focused on the glassiest portion of each sample (typically the surface rind). Low magnification (20×) images provided an overview of vesicle textures and other large-scale features. An additional eight to twelve images at higher magnifications (100×, 250×, or 500×) provided sufficiently detailed coverage for quantitative measurement of microlite textures. More images were acquired for samples in which crystal distribution was spatially heterogeneous, and higher magnifications were used for samples with smaller crystals. Imaged samples typically contain olivine ± plagioclase ± pyroxene (Fig. 4a). Areas of quench crystallization, identified by the presence of abundant small or dendritic crystals or pyroxene overgrowths on plagioclase microlites (Fig. 4b), were avoided during imaging. Plagioclase and pyroxene microlite crystallinities were measured directly from BSE images with the aid of automated image processing software (SCION Image). The large atomic number contrast between plagioclase and surrounding glass (Fig. 4) allowed automatic identification and measurement of this phase. Because the contrast between pyroxene and glass is less pronounced, pyroxene microlite crystallinities were determined by point counting BSE images with a transparent 690 point grid. Pyroxene could generally be distinguished from olivine by its mottled appearance, produced by patchy intergrowths of augite and pigeonite (Fig. 4c). Total microlite crystallinities (ϕm) are the sum of the area fractions of plagioclase and pyroxene in each image, normalized to exclude vesicles and phenocrysts. The accuracy of this method for determining ϕm is confirmed by previous studies, in which crystallinities estimated from BSE images are in good agreement with those calculated using mass balance techniques (Cashman et al., 1999; Hammer et al., 1999). Phenocryst contents were measured separately by a 1000 point count on an optical microscope, as the spatial distribution of olivine is heterogeneous at the scale of BSE imaging. Quantification of crystal number densities (number of crystals per unit area or volume) was limited to plagioclase, which comprises ~50% of the crystal content of most samples and is more easily imaged than pyroxene. Plagioclase also plays a key role in the rheological evolution of basaltic lava flows, as its tabular morphology promotes the development of a “touching framework” of microlites that contributes to the onset of yield strength and non-Newtonian behavior (e.g., Kerr and Lister, 1991; Saar et al., 2001; Soule and Cashman, 2005). Areal plagioclase microlite number densities (Na) were calculated from the number and area fraction of plagioclase microlites

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

145

Fig. 4. Backscattered electron (BSE) images of lava flow samples from the 1859 eruption. (a) Typical lava flow sample with phases labeled (plag = plagioclase; cpx = clinopyroxene). (b) Quench crystallization evidenced by abundant small and dendritic pyroxene crystals. Quench crystallization has variously affected samples from the 1859 eruption and was avoided during imaging and analysis. (c) Characteristic intergrowths of augite (light gray) and pigeonite (dark gray) in pyroxene microphenocrysts. (d) Skeletal olivine microphenocryst, indicating rapid growth. (e), (f) Radiating and plumose intergrowths of plagioclase and pyroxene, textures associated with high cooling rates.

and used to determine the average crystal size, d (= (ϕplag/Na)1/2). Two-dimensional measurements were then converted to volumetric number densities (Nv) according to the relationship Nv = Na/d (Underwood, 1970). For ease of comparison with other studies, we report both areal and volumetric number densities in Appendix 3 (supplementary material). 5. Results 5.1. Field observations and mapping Field- and aerial photo-based mapping support our interpretation of contemporary accounts of the 1859 eruption. Both the upper and north vents appear to represent early stages of activity, with small

constructive features and simple relationships between the vents and the flows they fed. In contrast, the lower vents are larger and more complex, consistent with their being the primary source of magma for both the main phase 'a'ā flows and the subsequent pāhoehoe flow field. The primary 'a'ā channel is striking in the length to which it is lined with pāhoehoe overflows (N35 km, longer than the length of most channel-fed Hawaiian lava flows; Fig. 3). In some locations within the saddle region, channels can be seen within (and below) overlying tube-fed flows. Pāhoehoe-lined channels are laterally surrounded by, and grade down-flow into, clinkery 'a'ā. These changes in surface morphology reflect a progression from the spreading (and clinker-forming) flow front to mature (and pāhoehoe-lined) channels, as documented by Lipman and Banks (1987) during the 1984 eruption of Mauna Loa. Tube-fed flows show evidence of overflows from

146

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

skylights in the low-slope saddle region and well-developed tumuli and inflation features on the coastal plain (e.g., Hon et al., 1994), consistent with historical accounts of protracted emplacement. 5.2. Bulk compositions Bulk compositions of down-flow samples, as well as those of spatter and proximal flow samples from each of the three vent areas, confirm the compositional heterogeneity noted in previous studies (Rhodes, 1983; Montierth, 1999). The compositions of medial and distal samples are distributed according to eruptive phase (Fig. 5a), with most channel-fed ('a'ā and overflow) samples having 8.7–10.1 wt.% MgO and most tube-fed (pāhoehoe) samples having 7.2–7.8 wt.% MgO. Three exceptions to the observed correlation between bulk composition and apparent eruptive phase – two low-MgO 'a'ā samples and one highMgO pāhoehoe sample – were collected from the saddle between Mauna Loa, Mauna Kea, and Hualālai volcanoes. In this region, the 'a'ā and pāhoehoe flow fields are complexly interfingered, and flows may have been fed by any of the three vent areas (Fig. 3). We therefore take the clear compositional bimodality of distal samples as evidence that

bulk composition corresponds to eruptive phase (channel-fed versus tube-fed), regardless of surface morphology. Compositions of samples from each of the three vent areas are also strongly clustered (Fig. 5b). Spatter and proximal flow samples from the lower vents have high-MgO bulk compositions identical to those of down-flow channel samples, confirming the source of channelforming lava inferred from mapping and narrative accounts. In contrast, samples from the upper and north vent areas have low-MgO bulk compositions similar to those of down-flow samples associated with the tube-fed phase of the eruption. Thus low-MgO 'a'ā samples from the saddle region may be remnants of the earliest phase of activity (erupted from the upper or north vents), while the lone high-MgO pāhoehoe sample may represent the waning stage of the 'a'ā phase, as described by Alexander (1859) during his visit to the lower vents on February 9–11. 5.3. Properties of proximal lavas The textures, phase assemblages, and quenching temperatures of proximal samples record eruption conditions at each of the three vent areas. All spatter samples contain numerous round vesicles and sparse olivine phenocrysts (0.5–3 mm) and microphenocrysts (0.1–0.5 mm). Total olivine phenocryst plus microphenocryst contents determined by point count for a suite of representative samples are variable and range from b0.1–5.4 vol.% in high-MgO samples and b0.1–1.6 vol.% in low-MgO samples (Appendix 3, supplementary material). Olivine typically displays hopper morphologies or elaborate skeletal textures (Fig. 4d) that reflect rapid growth at large undercoolings (Donaldson, 1976; Faure et al., 2003). Plagioclase and pyroxene are present only as small microphenocrysts (b0.2 mm in longest dimension) in spatter from the upper and north vents; these phases are absent from lower vent spatter. In proximal flow samples, radiating and plumose intergrowths of plagioclase and pyroxene (Fig. 4e, f) are common and indicate disequilibrium crystallization (e.g., Lesher et al., 1999). The quenching temperatures of spatter and proximal flow samples can be calculated from measured glass MgO contents (cf. Eq. (1); Table 1) and serve as a proxy for eruption temperatures. Spatter temperatures cluster by vent area (Fig. 5c) in a manner consistent with observed mineral assemblages. Glass in lower vent spatter samples, which contain only olivine crystals, records eruption temperatures ranging from 1194 °C to a maximum of 1216 °C. In contrast, glass in proximal samples from the upper and north vents yields cooler maximum eruption temperatures of 1166 °C and 1168 °C, respectively, consistent with their peritectic mineral assemblages (olivine + plagioclase + pyroxene). Additionally, several low-temperature north vent spatter samples (Table 1) contain abundant plagioclase and pyroxene microphenocrysts that may reflect syn- or pre-eruptive degassing (e.g., Lipman et al., 1985). Together, crystal phases and glass compositions demonstrate that early-erupted magma from the upper and north vents was similar to the ‘normal,’ shallowly-stored magmas that have erupted over most of Mauna Loa's recent history (the ‘reservoir lavas’ of Rhodes, 1995). The bulk compositions of later tube-fed pāhoehoe samples (Section 5.2) are also typical of historically-erupted magmas. Such magmas are characterized by compositional homogeneity (6.7–8.0 wt.% bulk MgO; e.g., Rhodes, 1983) and a tendency to erupt at, or very near, the temperature of multiple-saturation (e.g., Rhodes, 1995). In contrast, magma erupted from the lower vents was hotter and less evolved than typical rift zone compositions, but similar to melts that hosted the olivine-rich picrites erupted in 1852 and 1868 (Rhodes, 1995). 5.4. Changes in lava properties during transport

Fig. 5. Histograms showing the bimodal distribution of bulk and glass MgO in lava samples from the 1859 eruption. 2σ error in MgO (wt.%) is narrower than bin width. (a) Distribution of bulk MgO in lava flow samples. (b) Distribution of bulk MgO in vent spatter samples. (c) Distribution of glass MgO and quenching temperatures in spatter samples. Temperatures were calculated from glass compositions using Eq. (1).

5.4.1. Temperature Quenching temperatures of proximal and down-flow samples can be used to infer temperature–distance relationships in the channel-

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

and tube-fed flow fields. Lower vent spatter and pāhoehoe-surfaced overflows show that channel temperatures decreased systematically with increasing distance from the vent (Fig. 6a). Cooling was most rapid in the proximal channel reaches, as indicated by a drop in temperature from a maximum of 1216 °C to 1160 °C over the first 10 km of transport (5.6 °C/km). In the medial section of the channel (10–36 km), pāhoehoe overflows record a steady, less dramatic decrease from 1160 °C to 1146 °C (0.5 °C/km). Overflows are absent from the distal channel (N36 km; Fig. 3), and 'a'ā clinkers collected in this region were generally too crystalline for glass analysis. However, a single glassy 'a'ā sample at 45 km yields a quenching temperature of 1145 °C, suggesting that lava in the channel core was able to travel the last 10–15 km to the ocean with very little additional cooling. We have limited data to constrain cooling rates in the pāhoehoe flow field, as proximal lava was fed primarily through tubes during this phase of the eruption. In fact, our sampling suggests a near absence of spatter or surface flows accompanying the onset of tubefed activity, consistent with narrative accounts of rapid tube development that, by June of 1859 (3–4 months after the start of pāhoehoe production), was transporting lava to a distance of ~ 40 km from the source (Haskell, 1859b). To resolve temperature–distance relationships for tube-fed samples, we therefore assume that eruption temperatures were similar to those of the compositionally-identical upper and north vent lavas (1166 °C and 1168 °C, respectively). Medial samples with quenching temperatures of 1150–1158 °C thus record 8–18 °C of cooling over the first 20 km of transport (0.4–0.9 °C/km), substantially less than that in the proximal 'a'ā channel but within the range described

Fig. 6. Down-flow variations in quenching temperature and microlite crystallinity in lava samples from the 1859 eruption. Four proximal samples visibly affected by quench crystallization of pyroxene are not shown (see Table 1). (a) Quenching temperature versus distance from the lower vents. Error bars give uncertainty in the geothermometer (± 10 °C). (b) Volume fraction microlites (ϕm) versus distance from the lower vents. Error bars are 2σ (after Van der Plas and Tobi, 1965).

147

for other tube-fed flows (Peterson and Swanson, 1974; Cashman et al., 1994; Clague et al., 1999; Helz et al., 2003). Quenching temperatures of medial and distal tube-fed samples differ by up to 8 °C at equal sampling distances, probably reflecting small variations in both eruption and transport conditions. This temperature variation encompasses all of the variation observed in the pāhoehoe flow field, where glass temperatures never drop below 1149 °C (Fig. 6a). At distances N20 km, tube-fed samples have temperatures equal to or slightly higher than those of the (initially hotter) channel-fed samples at comparable transport distances (Fig. 6a), underscoring the thermal efficiency of tube-fed lava transport (as correctly inferred by Haskell, 1859b). 5.4.2. Crystallinity Both the channel and tube systems show an overall increase in total plagioclase and pyroxene microlite abundance (ϕm) with increasing transport distance (Figs. 6b, 7). However, the way in which that increase is achieved differs markedly between the two emplacement modes and provides insight into conditions of transport and cooling in lava channels and lava tubes. Crystallinity–distance relationships are best defined for glassy, pāhoehoe-surfaced channel overflows, where the volume fraction of microlites increases steadily from 0 to 0.25 over a distance of 36 km (0.007 ϕm/km). At greater distances, the absence of pāhoehoe overflows and paucity of well-quenched 'a'ā clinkers limits our ability to constrain distal crystallinity changes. However, the three glassy 'a'ā clinkers sampled in the distal channel have crystallinities of 0.25–0.30, consistent with their glass temperatures (Fig. 6a). If these samples are representative of core lava, it suggests that the crystallinity of lava within the channel remained approximately constant over the last 20 km of transport. Early stages of lava crystallization during tube-fed activity are difficult to quantify given the absence of proximal samples and the very different microlite contents of two medial samples at roughly equal transport distances (ϕm = 0.04 and 0.20 at ~ 20 km) (Fig. 6b). This discrepancy in crystallinity probably reflects the complexities of flow emplacement in the saddle region, where numerous flow branches may record both varied thermal histories (related to the extent of tube transport) and differences in initial eruption temperatures. Beyond the saddle region, microlite crystallinities of tube-fed pāhoehoe samples show little variation (ϕm = 0.17–0.23) between distances of 32 and 49 km. A direct comparison of channel- and tube-fed samples with pāhoehoe surface morphologies at ~35 km from the vent shows that both have comparable temperatures (c. 1150 °C) and microlite crystallinities (ϕm = 0.25). These samples differ, however, in how that crystallinity is achieved. While channel spillovers and 'a'ā clinkers show a dramatic down-flow increase in the number of plagioclase and pyroxene crystals, tube-fed samples show a pronounced increase in crystal size (compare Fig. 7a–c with d–f). Plagioclase microlite number densities (Na) highlight these differences in crystallization conditions: Na increases steadily down-flow in channel-fed lavas, but remains approximately constant in tube-fed lavas (Fig. 8a). As a result, the number densities of medial and distal channel-fed lavas are an order of magnitude higher than those of tube-fed lavas with comparable plagioclase microlite contents (Fig. 8b). Similar relationships between crystal size, crystal number density and lava type have been noted in previous textural studies of Hawaiian lava flows (Cashman et al., 1999; Katz and Cashman, 2003; Soule et al., 2004). 5.4.3. Vesicularity The vesicularity of both channel- and tube-fed lavas generally decreases with increasing transport distance (Figs. 7, 9). At equal distances from the vent, tube-fed samples are more vesicular than channel-fed 'a'ā samples, although channel spillover samples may have high vesicularities (Fig. 9). Similar trends are seen in samples collected along active lava tubes (e.g., Cashman et al., 1994) and lava

148

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

Fig. 7. BSE images showing down-flow changes in lava crystallinity. Scale bar is 200 μm in all images. (a) Lower vent spatter, 0 km. (b) Channel spillover, 28 km. (c) Channel 'a'ā, 40 km. (d) Upper vent spatter, 0 km (shown here to approximate proximal tube-fed lava; see text). (e) Tube-fed pāhoehoe, 21 km. (f) Tube-fed pāhoehoe, 40 km.

channels (Lipman and Banks, 1987; see Fig. 9 for comparison), in which vesicle content decreases rapidly away from the vent as bubbles are lost at the flow surface. This dramatic decrease in vesicularity causes a pronounced decrease in the volume of lava transported through a channel, a volume loss enhanced by actual mass loss to overflows and growing levees (e.g., Lipman and Banks, 1987; Baloga et al., 1998). Vesicularity variations also affect both the density and the rheology of lava transported through channel and tube systems, as confirmed by theoretical and analog models of lava flow behavior (e.g., Manga et al., 1998; Baloga et al., 2001; Rust and Manga, 2002; Llewellin and Manga, 2005). 6. Discussion A primary goal of this study is to use thermal and textural analysis of samples from solidified lava flows to interpret conditions of flow

emplacement. Such an approach extends our record of eruptive conditions beyond those observed in recent eruptions of Mauna Loa and Kīlauea (e.g., Crisp et al., 1994; Kauahikaua et al., 1998; Cashman et al., 1999; Helz et al., 2003; Kauahikaua et al., 2003), giving insight into the range of activity possible at Hawaiian volcanoes. This approach also has limitations: while spatial relationships between post-emplacement samples can be well constrained, temporal relationships must be inferred (e.g., Soule et al., 2004). Below, we show how bulk and glass compositions, quenching temperatures, and crystal textures of samples from the 1859 eruption may be used to draw qualitative and quantitative conclusions about the dynamics of flow emplacement, including details of eruption chronology, rates and mechanisms of cooling and crystallization, and the thermal and rheological evolution of the primary 'a'ā channel. We also discuss ways in which our data may be applied to predictive models of lava flow behavior, with particular emphasis on hazards posed by flank

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

149

shallow levels prior to eruption. We further suggest that pressurization and early eruption of this shallow reservoir magma may have been triggered by the ascent of primitive magma from depth, which later emerged at high effusion rates to form the lower vents and the 'a'ā flow field. Once buoyancy of the MgO-rich magma was lost or its supply was exhausted, low-MgO reservoir magma appears to have re-annexed the pathway established by channel-forming lava. Persistent effusion of reservoir magma over the following 10 months most likely occurred as the shallow system slowly depressurized, conditions fostered by rapid depressurization and the establishment of open pathways to the surface earlier in the eruption (e.g., Rowland and Walker, 1990). The unusual spatio-temporal variation in magma composition may be characteristic of paired eruptions initiating at radial vents, where ascending magma may perturb, but not necessarily mix with, magma stored in the shallow reservoir. 6.2. Cooling and crystallization of the 1859 lava flows

Fig. 8. Variations in plagioclase microlite number density in channel- and tube-fed lavas from the 1859 eruption. Symbols are as in Fig. 6, except where noted. (a) Areal plagioclase microlite number density (Na) versus distance from the lower vents. (b) Na versus plagioclase microlite crystallinity. Steep slopes indicate nucleation-dominated crystallization, while low slopes indicate growth-dominated crystallization. Error bars are 2σ, based on the standard deviation of separately analyzed SEM images of the same sample. In several cases images were not analyzed separately; for these samples, 2σ error is assumed to be similar to those of samples with similar Na.

eruptions. We end by placing the 1859 eruption in the context of the unusual activity at Mauna Loa during the last half of the 19th century. 6.1. Eruption chronology Measured bulk compositions link MgO-rich lavas to the early, channel-fed phase of the 1859 eruption that initiated at the lower vents. In contrast, low-MgO lavas were erupted first from the upper and north vents and later (during the long-lived tube-fed phase of the eruption) from the lower vents. Combining these observations with the chronology inferred from historic accounts and field observations, it appears that several compositional shifts occurred during the 1859 eruption: (1) a shift from more evolved to more primitive compositions as activity moved from the upper and north vents to the lower vents during the first few hours of eruption, and (2) several weeks later, a shift back to more evolved compositions accompanying the transition from channel- to tube-fed flow. Compositionally heterogeneous eruptions are rare in Mauna Loa's historic record and generally reflect variable olivine content in an otherwise homogenous melt (Rhodes, 1983). This is not the case for the 1859 eruption, where olivine content alone cannot account for variations between the two bulk compositions and differences in the erupted compositions of their respective host melts (Section 5.3). Thus, while these magmas may ultimately be derived from the same source, they have not evolved to the same degree. Rather, the clear separation of erupted compositions by vent area, and also in time, suggests that the 1859 eruption tapped two (or more) discrete magma bodies that experienced little or no interaction immediately prior to and during eruption. Early-erupted magma is compositionally similar to ‘normal’ summit reservoir magmas and was probably stored at

Our data show that lava traveling through the 1859 channel cooled and crystallized substantially (Fig. 6). Syn-emplacement cooling and crystallization is a common feature of channel-fed lava flows in Hawai'i (e.g., Crisp et al., 1994; Cashman et al., 1999; Soule et al., 2004). In this way Hawaiian lava flows differ from those at Mt. Etna, which are erupted at low temperatures (b1080 °C; ~ 50% crystallinity; e.g., Pompilio et al., 1998) and show little evidence of cooling and crystallization during transport (Calvari et al., 1994). Here we examine the relationship between eruption temperature (a function of both bulk composition and pre-eruption conditions of magma storage and ascent) and post-eruption cooling and crystallization, particularly as it contributes to the ultimate distance traveled by a lava flow (e.g., Soule et al., 2004). 6.2.1. Eruption temperature Spatter temperatures inferred from glass geothermometry of samples from the 1859 eruption range from 1135 °C to 1216 °C (Table 1); the high calculated temperatures of lower vent samples corroborate the ‘white hot’ descriptions of proximal channel-fed lava (Alexander, 1859; Haskell, 1859a). To place our temperature data within a broader framework, we have compiled published glass analyses of subaerial Mauna Loa lavas and converted glass compositions to temperatures using Eq. (1) (Fig. 10). Temperatures of known proximal samples best approximate eruption temperatures. Among these, three groups emerge. The most abundant (1155–1170 °C) includes eruption temperatures of spatter from the upper and north vents, as well as early-erupted spatter from

Fig. 9. Vesicularity versus distance in lavas from the 1859 eruption. Symbols are as in Fig. 6. Shaded area represents data from the 1984 eruption of Mauna Loa (Crisp et al., 1994). Although the data are scattered, lavas from both eruptions show a trend of decreasing vesicularity with distance. Channel overflows from the 1859 eruption tend to have higher vesicularities than other samples (i.e., tube-fed pāhoehoe and 'a'ā clinkers) collected at similar distances from the vent.

150

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

Fig. 10. Histogram of quenching temperatures of subaerial Mauna Loa lavas, calculated from glass compositions using the geothermometer of Montierth et al. (1995). Known proximal samples are assumed to approximate eruption temperatures. Maximum quenching temperatures for the 1859, 1984, 1950, 1868, and 1852 eruptions are labeled. Temperatures above 1175 °C are rare; only picrites and Mg-rich lavas, including spatter from the 1859 eruption, record such high temperatures. Data are from this study, Rhodes (1988, 1995), Wilkinson and Hensel (1988), Crisp et al. (1994), Helz et al. (1995), Garcia (1996), Montierth (1999), Davis et al. (2003), C. Thornber (unpublished data), and P. Wallace (unpublished data). Uncertainty in the geothermometer is ±10 °C.

Mauna Loa's most recent eruption in 1984 (1168 °C; based on glass compositions given by Helz et al., 1995). Experimental and thermodynamic models show that most tholeiitic Mauna Loa magmas become multiply-saturated within this temperature range (Ghiorso and Sack, 1995; Montierth et al., 1995), which we take to represent equilibrium eruption temperatures of the ‘normal’ magma compositions described by Rhodes (1995). A second, less-abundant cluster of eruption temperatures at 1135–1150 °C is lower than expected for near-peritectic Mauna Loa compositions. Low temperatures may indicate either poor quenching or syn-eruptive crystallization in response to degassing, as seen in vent lavas erupted from Mauna Loa in 1984 (e.g., Lipman et al., 1985; Crisp et al., 1994; Helz et al., 1995; Montierth et al., 1995). Several well-quenched north vent spatter samples from the 1859 eruption also fall within this temperature range. These samples have a wide range of crystallinities (ϕm = 0.02–0.35; Table 1) that support the interpretation that low eruption temperatures may record degassing-induced crystallization prior to or during eruption. The final group (T = 1175–1216 °C) includes maximum eruption temperatures of samples from infrequent eruptions of picrite and Mg-rich lava (from 1859, 1950, 1852, 1868; Rhodes 1983). Among these samples, the highest calculated temperature (1216 °C) is that of spatter from the lower 1859 vents, which appears to exceed even those of the 1852 and 1868 picrites, although no detailed work has been done on the glass chemistry of vent samples from these eruptions. Regardless, the uniformly high quenching temperatures of lower vent spatter (1194–1216 °C; Section 5.3) are clearly atypical of Mauna Loa lavas. If eruption temperatures at Mauna Loa tend to fall within a narrow range, then the effect of starting temperature on down-flow cooling and crystallization is limited. Perhaps for this reason, variations in eruption temperature have not been thoroughly explored in comparative studies of Hawaiian lava flows (Walker, 1973; Malin, 1980; Rowland and Walker, 1990; Pinkerton and Wilson, 1994; Rowland et al., 2005). The varied and unusual eruption temperatures of the 1859 lavas, however, lead us to examine the effect of eruption temperature on cooling and crystallization of the channel- and tube-fed flows. 6.2.2. Cooling and crystallization The extent of crystallization accompanying cooling of a given composition can be modeled using MELTS (Ghiorso and Sack, 1995)

and compared with our data from channel- and tube-fed samples (Fig. 11). The measured and modeled data show a good correspondence and illustrate the profound differences in initial cooling and crystallization paths experienced by high- and low-MgO magmas. At its maximum eruption temperature of 1216 °C, the MgO-rich lower vent lava (source of the early, channel-fed activity) is saturated only with olivine and must cool by 46 °C before reaching the plagioclase liquidus. Over this temperature interval (1216–1170 °C), the lava crystallizes just 4 vol.% olivine (dϕ/dT ≈0.001 ϕ/°C, where ϕ is the total volume fraction crystals). In contrast, at its estimated eruption temperature of 1166–1168 °C, the low-MgO 1859 lava must cool only slightly before the onset of plagioclase and pyroxene crystallization. Once the lava is multiply-saturated, temperature is buffered by the latent heat of crystallization, and dϕ/dT increases by an order of magnitude (to ~0.01 ϕ/°C). These observations demonstrate the importance of crystallization assemblage in determining cooling rate. At a constant rate of heat loss, single-phase crystallization will result in larger temperature decreases than multi-phase crystallization (e.g., Helz et al., 2003), reflecting the relatively small amount of latent heat generated in the olivine + liquid system. Thus, all other variables being equal, the high eruption temperature of early channel-forming lava would have permitted rapid cooling (with minimal crystallization) until the temperature reached the peritectic, at which point both lava compositions would have cooled and crystallized at the same rate. Our field data support this inference. Examination of down-flow samples shows that the substantial temperature drop observed in the proximal 'a'ā channel (a maximum of 5.6 °C/km between 0 and 10 km; Fig. 6a) corresponds to the regime of rapid cooling accompanied by minor olivine crystallization predicted by phase equilibria (Fig. 11). An overflow sample at 10 km with ϕm = 0.08 indicates that channel lava became multiply-saturated somewhere between 0 and 10 km from the vent, after which the buffering effects of latent heat helped to reduce cooling rates to ~0.5 °C/km along the medial and distal channel. These cooling rates are similar to those of tube-fed samples from the 1859 eruption (0.4–0.9 °C/km between 0 and 20 km) and from recent tube-fed eruptions of Kīlauea

Fig. 11. Modeled and measured temperature versus crystallinity in lavas from the 1859 eruption. Solid lines show calculated 1-bar equilibrium crystallization of average highMgO and low-MgO bulk compositions, assuming 0.1 wt.% H2O and FMQ buffer conditions (Ghiorso and Sack, 1995). We assume an fO2 of FMQ because models using NNO produced orthopyroxene (which is not observed) instead of high- and low-Ca clinopyroxene phases (which are observed). Volume fraction crystals is calculated as microlite crystallinity (ϕm) plus the modeled volume fraction of olivine. For samples containing olivine only, volume fraction crystals is the amount of olivine measured by point count (Appendix 3, supplementary materials). Approximate dϕ/dT values are based on linear fits to modeled data between inflection points. Phases crystallizing in each temperature interval (± 5 °C) are labeled. Error bars are 2σ.

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

(0.4–1.0 °C/km, e.g., Clague et al., 1995; Cashman et al., 1999; Helz et al., 2003). Changes in temperature and crystallinity with distance may be converted to temporal cooling and crystallization rates (°C/s and ϕm/s) if flow velocities are known (e.g., Crisp et al., 1994; Cashman et al., 1999; Soule et al., 2004). Such data are useful if we wish to draw direct comparisons between flows and constrain parameters used to model flow behavior. Based on observations by Haskell (1859a) of high velocities at the vent (“such a rate that the eye could scarcely follow it”) and estimates of medial channel velocities of less than 4–5 m/s, we feel that velocities of 4–6 m/s in the proximal channel and 1–2 m/s in the saddle region provide conservative estimates of cooling and crystallization rates. Our calculations suggest that cooling was rapid in the proximal channel: 0.01–0.03 °C/s, or 36–108 °C/h, between 0 and 10 km. These rates are 2–6 times higher than those reported in previous studies (~0.005 °C/s; e.g., Crisp et al., 1994; Cashman et al., 1999; Soule et al., 2004) and are clearly not sustainable over the length of the channel. Rapid cooling is most likely a consequence of single-phase crystallization, promoting rapid temperature decrease, and high near-vent velocities, which prevent formation of an insulating surface crust (e.g., Cashman et al., 2006), thereby increasing rates of radiative heat loss. Despite high proximal cooling rates, proximal crystallization rates (0.3–0.5 × 10− 4 ϕm/s) are somewhat lower than those reported elsewhere for active lava channels (0.4–1.0 × 10− 4 ϕm/s in the proximal 1984 channel and 0.5–0.8 × 10− 4 ϕm/s in a proximal Kīlauean channel; Crisp et al., 1994; Cashman et al., 1999), underscoring the significance of high eruption temperatures. Estimated cooling rates of b0.001 °C/s (b3.6 °C/h) through the saddle (10–36 km) are an order of magnitude lower than those estimated for the proximal channel; crystallization rates are also low (0.1–0.2 × 10− 4 ϕm/s). Low rates of cooling and crystallization point to the buffering effects of latent heat once the peritectic is reached, although comparison with crystallization rates estimated for active 'a'ā channels (~0.5 × 10− 4 ϕm/s; Crisp et al., 1994; Cashman et al., 1999) suggests that our medial flow velocity estimates may be overly conservative. The only velocity estimate available for the tube-fed phase of the eruption is an ocean-entry observation of 2–3 miles/h (0.9–1.3 m/s) on June 22 (Haskell, 1859b). This is similar to average flow rates measured in active Hawaiian lava tubes (1.5–3 m/s; Peterson et al., 1994; Kauahikaua et al., 1998) and yields cooling rates of 0.001–0.003 °C/s for the 1859 pāhoehoe flow. These cooling rates are similar to both recent estimates for tube-transported lava at Kīlauea (0.001–0.002 °C/s; Cashman et al., 1999) and rates inferred for the medial and distal 1859 channel (above). Thus the low inferred rates of cooling and crystallization in the medial to distal channel may result, in part, from increased insulation of the flow interior as the channel developed a mature crust (e.g., Crisp and Baloga, 1990; Cashman et al., 1999; Griffiths et al., 2003). Once thermally stratified, open-channel flows cool slowly and may lengthen considerably before crystallinity is sufficient to inhibit viscous flow. Taken together, our rate estimates suggest that specific parameters used in lava flow models (e.g., cooling rates, crystallization rates, and contributions from latent heating), currently calibrated for multiplysaturated compositions, may not apply to lava that erupts at high temperatures. Our data also stress the importance of incorporating changes in the thermal structure of lava channels into flow models (e.g., Crisp and Baloga 1990; Griffiths et al., 2003; Cashman et al., 2006). 6.2.3. Mechanisms of crystallization Textural data indicate that crystallization proceeded differently during the channel- and tube-fed phases of the 1859 eruption (e.g., Fig. 8), with a dramatic increase in crystal number (Na) along the primary channel contrasting with nearly constant Na along the pāhoehoe flow field. These distinct crystallization regimes may be

151

described as nucleation- and growth-dominated, respectively, and have been attributed to different rates of radiative cooling in open channels as compared to well-insulated lava tubes (Cashman et al., 1994; Crisp et al., 1994; Helz et al., 1995; Cashman et al., 1999; Polacci et al., 1999; Helz et al., 2003; Katz and Cashman, 2003; Soule et al., 2004). Viewed in this context, nucleation-dominated crystallization in the primary channel reflects rapid radiative cooling of hightemperature lava in the channel core, a process associated with high flow rates, efficient mixing, and strong sidewall shear (e.g., Griffiths et al., 2003; Cashman et al., 2006). Conversely, the growth-dominated textures of tube-fed samples record slow cooling via conduction through a continuous crust. Importantly, the textural characteristics of channel-fed samples are similar regardless of flow morphology (i.e., 'a'ā clinker or pāhoehoe overflow; Fig. 8). The textural distinctions between channel- and tube-fed lavas can further be used to address the origin of samples collected in the complex saddle region. For example, we expect that the lone highMgO pāhoehoe sample (901-6p) was associated with channel-fed activity because of its bulk composition (Section 5.2). This hypothesis is confirmed by the measured Na (N400/mm2), which is similar to that of a channel overflow sampled at an equal transport distance (Fig. 8a). We interpret this sample to be a pāhoehoe breakout from the primary channel that preserves the plagioclase nucleation characteristics of lava at the point of breakout; its slightly higher plagioclase crystallinity (0.08, versus 0.05 in the corresponding channel overflow; Fig. 8b) may record continued crystal growth prior to quenching. Syn-eruptive crystallization driven by degassing also appears to have played a role in cooling and crystallization of the 1859 lavas. The wide range of glass compositions in quenched spatter samples (5.4–6.8 wt.% MgO at the north vents and 7.9–8.9 wt.% MgO at the lower vents) documents rapid crystal growth prior to (or during) eruption. North vent samples, in particular, record N30 vol.% crystallization. Examination of individual samples shows that the measured crystal abundance correlates well with changing glass temperatures (Fig. 12). Additionally, these crystals are distinct from those formed during down-flow cooling: they are larger and of lower number density, such that they form a discrete microphenocryst population. Similar patterns of crystallization observed during the 1984 eruption (~20 vol.% microphenocryst crystallization at the vent during the first 100 h; Crisp et al., 1994) were interpreted to reflect an approach to stable conditions following initial disequilibrium crystallization upon gas loss (Lipman et al., 1985; Crisp et al., 1994).

Fig. 12. Quenching temperature versus crystallinity in north vent spatter samples from the 1859 eruption. Scale bar is 100 μm in all images. The microphenocryst content of the spatter samples varies widely (ϕ = 0.02–0.35), suggesting that subsurface crystallization, perhaps facilitated by degassing, occurred after the onset of eruption. Error bars are 2σ.

152

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

Syn-eruptive, degassing-induced crystallization may in fact be common in Hawaiian eruptions (e.g., Lipman and Banks, 1985; Helz et al., 1995; Katz and Cashman, 2003; Fig. 10). Furthermore, the rheological consequences of syn-eruptive crystallization could be profound. Specifically, if viscous flow ceases at a maximum crystal packing fraction (ϕmax) of ~0.5–0.6 (e.g., Marsh, 1981), crystallization equivalent to that observed at the north vent would account for more than half of the crystallization required to cause flow to cease (e.g., Pinkerton and Wilson, 1994). In contrast, the net effect of degassing on the rheology of the high-MgO 1859 magma would have been negligible due to the large amount of cooling required to produce substantial crystallization (Fig. 11). 6.3. Evolution of the 1859 channel 6.3.1. Rheological evolution of the primary channel Changes in crystallinity and vesicularity along the primary channel would have affected the rheological properties of channelized lava, in turn affecting conditions of flow emplacement. The effect of bubbles on suspension rheology, a function of vesicle content, bubble size, and shear rate, is difficult to quantify; however, its effect on lava viscosity was likely less than an order of magnitude (e.g., Rust and Manga, 2002; Pal, 2003). We therefore restrict our rheological examination to the effect of crystallization on lava in the primary channel. Modeling the rheology of particle-melt suspensions is complex, as particle interactions are controlled by particle shape, size distribution, concentration, and flow regime (Stickel and Powell, 2005). Most important is the value of maximum particle packing (ϕmax), which dictates the ‘locking point’ of the suspension and depends on the shapes and sizes of suspended particles, as well as the imposed stress (e.g., Wildemuth and Williams, 1984; Barnes et al., 1989). To account for variations in crystal morphologies observed in Hawaiian lavas, we calculate the reduced viscosity (ηr = ηs/ηo, where ηs is the suspension viscosity and ηo is the liquid viscosity) using a version of the Krieger– Dougherty equation modified to account for a bimodal population of suspended particles: h i h i − ½η/max − ½η/max ð1− ð/ =/max ÞÞ ηr = ð1 − ð/ =/max ÞÞ pl

px

ð2Þ

(Barnes et al., 1989). Values of shape factor [η] and ϕmax for plagioclase (pl) and pyroxene (px) are approximated using experimental data for tabular and spherical particles, respectively (Barnes et al., 1989). Liquid viscosities (ηo) were calculated from glass compositions and quenching temperatures using the empirical model of Shaw (1972). We do not consider olivine crystals in our bulk viscosity calculations because the volume fraction of olivine is small (b0.05) and because fine particles (i.e., microlites) play a primary role in determining the rheological properties of suspensions with distributed particle sizes (e.g., Probstein et al., 1994). Our calculations suggest that the viscosity of channelized lava (Fig. 13a) increased slowly along the first 20 km of transport and then more rapidly in the distal reaches of the channel. The very high crystallinities (and effective viscosities) of both channel overflows and 'a'ā clinkers at ~35 km suggest that lava deformation in the distal channel was no longer in the viscous regime (e.g., Soule and Cashman, 2005). Also at this distance, pāhoehoe spillovers are absent from the primary channel and the channel itself becomes poorly-defined. Thus the breakdown of the channel and its smooth-surfaced linings may be a direct consequence of a shift from Newtonian to non-Newtonian behavior as lava crystallinity increased beyond a critical value. For comparison, we used a similar approach to calculate the viscosity of down-flow channel samples from the 1984 eruption (using data from Crisp et al., 1994; Fig. 13a). We have excluded microphenocrysts from these calculations in an effort to directly compare the affects of syn-emplacement microlite crystallization on the

Fig. 13. Viscosity and advance rate estimates for channel-fed 1859 lava. (a) Bulk viscosity versus distance from the vent in lavas from the 1859 and 1984 eruptions. See Section 6.3.1 for a detailed description of viscosity calculations. Shaded areas are visual fits to the viscosity data. Values of [η] and ϕmax for plagioclase and pyroxene are [η]plag = 9.87, [η]pxn = 3.28, ϕmax (plag) = 0.382, ϕmax (pxn) = 0.61 (after Barnes et al., 1989); all calculations assume anhydrous lavas. Error in viscosities is roughly an order of magnitude (see Crisp et al., 1994). (b) Time versus distance traveled for recent channelfed Hawaiian eruptions (modified from Kauahikaua et al., 2003). Data are grouped according to effusion rate.

rheology of the two flows (including microphenocrysts would further increase our viscosity estimates). This comparison highlights the remarkable fluidity of lava in the 1859 channel to great distances, in marked contrast to the 1984 eruption, where rapid crystallization at the vent and along the channel caused substantial increases in viscosity within several kilometers of the source (Moore, 1987). The low viscosity of the 1859 lava would also have affected the rate of flow advance, particularly in proximal locations. In Fig. 13b, we compare the inferred advance rate of the 1859 'a'ā flow, based on the known constraint of 8 days travel time from the vent to the coast (Dana, 1859), with rates of flow advance for recent channel-fed eruptions of Kīlauea and Mauna Loa (Kauahikaua et al., 2003). We assume that the flow decelerated with distance; this is a reasonable assumption for basaltic flows not limited by lava supply (Kilburn, 1996), reflecting the tendency of flows to slow and thicken in response to down-flow viscosity and density increases (e.g., Baloga et al., 2001). Our comparison suggests that the 1859 'a'ā flow advanced rapidly, even for its high effusion rate (which we here take to be ~ 400 m3/s for the time period of interest; Section 2). 6.3.2. Length of the primary channel It is generally agreed that effusion rate is the dominant factor controlling both the lengths and advance rates of channel-fed flows (e.g., Walker, 1973; Pinkerton and Wilson, 1994; Kauahikaua et al., 2003). This correlation reflects a limit to temporal cooling rates, which

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

allows fast flows to advance farther before cooling to limiting temperatures. However, effusion rate is clearly not the only parameter controlling the length of Hawaiian lava flows, as illustrated by the data summarized in Fig. 2 and highlighted by comparison of the 1859 and 1984 eruptions. Both eruptions produced channelized flows of similar volume (220 × 106 m3 in 1984 versus 270 × 106 m3 in 1859) and duration (20 days in 1984 versus 18–27 days in 1859), but the 1984 flow traveled only 27 km (Lipman and Banks, 1987), half the length of the 1859 flow (which was terminated by reaching the ocean). Although similar in many respects, we have shown that the 1859 and 1984 eruptions represent rheological endmembers (Section 6.3.1). As such, they demonstrate the degree to which the temperature and crystallinity of erupted lava may influence flow length. Lava in the 1859 channel erupted at remarkably high temperatures (1194–1216 °C) to produce nearly aphyric lavas that remained fluid over long distances (Fig. 13). This is despite rapid cooling in proximal regions and up to 22 °C cooling at the vent (Section 6.2.2). In contrast, the five-fold increase in apparent viscosity observed during the 1984 eruption (Moore,1987) can largely be attributed to a combination of syn-eruptive microphenocryst crystallization and down-flow microlite crystallization (Crisp and Baloga, 1994). Interestingly, of the 30 channelized Hawaiian lava flows plotted in Fig. 2, only the 1859 channel is longer than 27 km (although the lengths of some of the highest effusion rate flows are minima because they reached the ocean). If the low modal quenching temperatures of many proximal Mauna Loa lavas (Fig. 10) record degassinginduced crystallization, perhaps the tendency of some flows to stop well before Walker's upper limit reflects this rheological constraint. Based on these data, we suggest that the physical state of magma at the vent (temperature, crystallinity, and effective undercooling) plays an important role in modulating the length of Hawaiian lava flows. Syn-eruptive crystallization may in turn affect magma flux. Lipman and Banks (1987) described an inverse relationship between effusion rate and the microphenocryst content of vent spatter during the 1984 eruption, suggesting that decreasing eruption rates were linked to the changing rheology of erupted magma. Additionally, down-flow changes in bulk viscosity triggered stagnation that proceeded upslope from the flow toe (similar to the ‘transitional regime’ recognized in analog experiments; Griffiths et al., 2003; Cashman et al., 2006), a process amplified by diminishing lava production at the vent (Lipman and Banks, 1987). Thus the low viscosity and Newtonian rheology of lava erupted throughout the channel-fed phase of the 1859 eruption may actually have helped to sustain the high effusion rates that contributed to the channel's unusual length. Flow length is also affected by slope traversed. In particular, openchannel flows slow and thicken on lower slopes, which exert less gravitational stress on flow crust than steep slopes, promoting crustal coverage. Furthermore, flow thickening reduces the ratio of exposed surface area to lava volume (Pinkerton and Wilson, 1994) and changes the internal convection regime (Griffiths et al., 2003). Both processes decrease rates of flow cooling. Maximum flow length (Lm) has therefore been related to the inverse of mean ground slope as 2

Lm = 2:53λ = δch

ð3Þ

where λ is 1/sin β (β is the underlying slope) and dch is 0.008 m, the thickness of chilled crust after a characteristic chilling time (Kilburn 2004). Applying this relationship to the 51 km long 1859 channel predicts an average slope of b4.5°. We examine this prediction using slope data from Rowland and Garbeil (2000). Starting at the vents, slopes gradually decrease from a maximum of 9° near the summit to b5° at lower elevations on Mauna Loa's west flank. The most dramatic topographic feature along the flow path is the saddle between Mauna Loa, Mauna Kea, and Hualālai volcanoes, where average slopes are just 2–3° (Fig. 3). An approximate flow path modeled using FLOWGO (Harris and Rowland, 2001) shows the average slope underlying the 1859 channel to be ~4°, in agreement

153

with Eq. (3). However, an equivalent analysis would indicate that the 1984 eruption should have traveled farther than its 27 km length. As discussed above, the length of this flow may have been limited, in part, by syn-eruptive crystallization, and accompanying viscosity increases, at the vent. This simple analysis ignores the complexity of real lava flows, for which thickness depends not just on slope, but also on mass loss to levees, volume loss (or gain) by degassing or inflation, and rheological changes during transport (e.g., Baloga et al., 1998; Baloga et al., 2001). Moreover, lava flows rarely remain confined to a single channel and may form complex branching patterns in response to both small topographic barriers and temporary blockages (e.g., Lipman and Banks, 1987; Cashman et al., 2006). Lipman and Banks (1987) cite a breakdown in the hydraulic efficiency of the 1984 channel system as a major cause of flow stagnation, forcing blockages and re-establishment of major tributaries well-upslope of the flow front. Historic accounts of the 1859 eruption also document a complex channel geometry at high elevations on the northwest flank (“a width in some places of three or four miles presented the appearance of shining rosecolored lace work”; Sleeper, 1859). Across the low-sloped saddle, extensive flow branching expanded the width of the 'a'ā flow field (Fig. 3; Section 2), concurrent with observations of local flow stagnation and ‘large drifting masses’ at channel surfaces (Alexander, 1859). This scenario conflicts with model predictions of increased flow thickening, and attendant insulation, on low slopes, and should have the opposite effect on flow cooling. Unfortunately, it is difficult to evaluate the effects of these conditions on the thermal evolution of the 1859 channel using field data, as 'a'ā within the saddle region is largely covered by later pāhoehoe flows. In contrast, below the saddle region, the 1859 'a'ā flow formed a remarkably straight, well-defined, mostly pāhoehoe-lined channel (Fig. 3). These observations raise interesting questions about the relationship between channel stability and flow length. In addition to topographic obstacles (Cashman et al., 2006) and blockages (Lipman and Banks, 1987), fluctuating effusion rates have been linked to the breakdown and widening of 'a'ā flow fields at Mount Etna (Guest et al., 1987; Calvari et al., 2002). Conversely, steady lava supply promotes the formation of narrower flow fields (Calvari et al., 2002). Channel geometry may also be important: spalling along channel meanders was a common source of large lava debris during the 1984 eruption (Lipman and Banks, 1987), and changes in channel geometry (i.e., bends, expansions, and contractions) can disrupt flow crust, affecting rates of radiative heat loss (Cashman et al., 2006). Straight, stable channels might instead be expected to enable efficient delivery of material from the vent to the flow front. Clearly, all of these factors may play a role in determining the lengths and heat budgets of lava channels, and they are grounds for continued study. 6.3.3. Applications to models of lava flow behavior Models of lava flow behavior aim to forecast the paths, lengths, advance rates, and areal extents of lava flows. Approaches to modeling flow behavior vary considerably, from simple empirical relationships between physical parameters (e.g., length, effusion rate, slope, duration, and channel dimensions; Walker, 1973; Pinkerton and Wilson, 1994) to probabilistic assessment of lava flow inundation hazard (Kauahikaua et al., 1995). Still other models describe open-channel flow in terms of thermal budgets (Crisp and Baloga, 1990, 1994), bulk rheology (Hulme, 1974; Tallarico and Dragoni, 2000), and crustal strength (Kilburn, 2004). FLOWGO (Harris and Rowland, 2001) models open-channel flow by incrementally advancing lava over DEM topography until loss of momentum, core solidification, or yield strength prevents further advance. By calibrating the model with field data from recent Hawaiian eruptions, Rowland et al. (2005) have used FLOWGO to assess hazards from channel-fed lava flows at Mauna Loa. Because FLOWGO models changes in thermal and rheological parameters,

154

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

permits a flexible range of input conditions, and is easy to run, it presents an excellent opportunity for comparison with our field data. We have modeled the 1859 channel using a range of input parameters. Our first run utilized inputs based on the well-documented 1984 eruption (after Harris and Rowland, 2001), as approximated by the model's default parameter file. Additional runs incorporated thermal and rheological properties consistent with the compositions of field samples (i.e., eruption temperatures of ~ 1200 °C, initial viscosities of 10–1000 Pa s, 1–2 vol.% phenocrysts, and 8–40% average vesicularity). For consistency with Rowland et al. (2005), we set the effusion rate of our runs to ~400 m3/s using input channel dimensions, assuming a square channel cross-section. FLOWGO also permits several modes of crustal growth that determine the way in which the fraction of crustal coverage, fcrust, changes with flow advance. Field studies and models suggest that fcrust plays an important role in modulating the heat budgets of lava flows (e.g., Crisp and Baloga, 1990; Kilburn, 1993; Hon et al., 1994; Cashman et al., 2006), and the model is highly sensitive to choice of crustal growth model (Harris and Rowland, 2001). In all cases, it was necessary to select a heavy crustal growth model (thus limiting cooling during flow) to achieve desired flow lengths at reasonable effusion rates (b~ 900 m3/s). The results of our model runs are shown together with our field data in Fig. 14. Although most runs are able to reproduce the observed flow length of ~ 50 km, modeled patterns of cooling and crystallization differ substantially from those given by channel samples. Most notably, FLOWGO overestimates proximal temperatures (and underestimates microlite crystallinities) in the 1859 channel (Fig. 14a, b), suggesting its cooling models are not calibrated for high-temperature lavas. These discrepancies persist when viscosity, effusion rate, and vesicularity are varied within reasonable limits. Inconsistencies with field data can be partly attributed to FLOWGO's default crystallization model, which is linked to a MELTS-based look-up table that assumes an eruption temperature of 1140 °C (approximating eruption temperatures measured in the field during recent eruptions of Mauna Loa and Kīlauea). Above 1140 °C, the model applies a constant temperature–crystallinity relationship (A. Harris, pers. comm., Oct. 2008); it therefore varies crystallinity as a linear function of temperature across the peritectic (compare our MELTS-modeled crystallization with FLOWGO's crystallization model; Fig. 14c). Additionally, recent analog studies have shown that (1) internal convection plays an important role in the cooling of open-channel flows, and (2) while crustal coverage (fcrust) is a predictable function of lava properties and flow dynamics in uniform channels, crust may rearrange rapidly in response to variations in channel geometry (Griffiths et al., 2003; Cashman et al., 2006). Our work indicates that accurate prediction of down-flow cooling and crystallization requires direct linkage of thermodynamic models (e.g., MELTS) and fluid dynamic models across a range of initial conditions. Ideally, predictive models should also incorporate the thermal effects of changing channel geometries, local controls on crustal development and disruption (e.g., Kerr et al., 2006), and stochastic models to mimic flow branching. Finally, further calibration with field studies is required to enhance our ability to apply models to hazard assessment. 7. Implications for hazards posed by flank eruptions at Mauna Loa The northwest flank of Mauna Loa is home to the overwhelming majority of the volcano's radial vents, only four of which have been active during the past 200 years (1843, 1877, 1859, 1935; Lockwood and Lipman, 1987). Among historic eruptions, the 1859 eruption is unique in erupting primarily (and perhaps exclusively) from subaerial radial vents. The 1859 eruption also occurred during a period of relative unrest (1843–1877; Lockwood and Lipman, 1987) marked by several unusual events, including the repeated eruption of primitive and picritic lavas closely spaced in time (1852, 1855, 1859, 1868;

Fig. 14. Comparison of FLOWGO model results and field data from the channel-fed phase of the 1859 eruption. Input parameters are discussed in Section 6.3.3. Shaded regions show range of model results when input viscosity, vesicularity, and starting channel dimensions are varied within reasonable limits (10–1000 Pa·s, 8–40%, and 4.0–4.5 m, respectively). (a) Temperature versus distance traveled. (b) Volume fraction microlites versus distance traveled. (c) Temperature versus volume fraction microlites.

Rhodes and Hart, 1995), a magnitude 7.5 earthquake in 1868 (Tilling et al., 1987), and average eruption rates twice as high as in preceding and subsequent decades (~ 47 × 106 m3/yr versus ~ 21 × 106 m3/yr since 1877; Lockwood and Lipman 1987). A steady, systematic decrease in the incompatible element abundances of erupted lavas during this period of high flux has been interpreted to reflect high rates of magma supply from a depleted parental magma source (Rhodes and Hart, 1995).

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

The off-axis vent location of the 1859 eruption could be a direct consequence of high supply rates if the northwest flank were unable to accommodate increased magma input by outward deformation (e.g., Lipman, 1980). High rates of supply would also encourage eruption of primitive (and high-temperature) compositions if rapid depletion of a compositionally stratified plumbing system allowed primitive magmas to ascend to unusually shallow levels (Rhodes, 1995). Alternatively, edifice pressurization may have permitted primitive magma stored at depth access to new or seldom-used transport paths. Taken together, these observations suggest that the likelihood of both radial vent eruptions and eruptions of Mg-rich lava increases during periods of high magma supply. As such, radial vent eruptions, while rare, may tap primitive magmas with proportionately greater frequency than summit and rift zone eruptions. Vent location may also have affected the crystal content of erupted magma. Rare subaerial eruptions of primitive lava at Mauna Loa have produced picrites rich in olivine phenocrysts (up to 10% in 1852 and up to 27% in 1868; Rhodes, 1995), inferred to have been entrained from rift zone magma reservoirs (Clague et al., 1995; Rhodes, 1995). Despite having a melt composition similar to these historic picrites, lava feeding the 1859 'a'ā channel was crystal-poor, suggesting that rising magma circumvented zones of cumulate olivine during subsurface transport. Thus the combined attributes of its off-axis (non rift zone) eruption location and its weakly-phyric character may be related. We end by noting that the eruption of primitive lava at flank, rather than rift zone, locations could pose unique hazards. Hot, rapidly erupted flows may advance quickly and cover great distances during the early stages of eruption (Fig. 13b). Unlike magmas passing through established rift zones (e.g., the 1852 and 1868 picrites), primitive lavas erupted on the northwest flank may be less likely to contain entrained or accumulated olivine, further contributing to low eruptive viscosities. Existing models of lava flow behavior underestimate both advance rates and total distance traveled for these initial conditions. For radial vents formed at low, rather than high, elevations on the volcano, these effects would be compounded by increased proximity to populated areas. 8. Conclusions Careful mapping, sampling, and compositional analysis constrain eruption and flow emplacement conditions pertinent to the 1859 eruption of Mauna Loa Volcano. The unusual length of its primary lava channel (51 km), and the extent to which the channel is lined with pāhoehoe overflows (36 km), was fostered not only by high effusion rates (e.g., Walker, 1973), but also by high eruption temperatures (1194–1216 °C), which delayed the rheological transitions that ultimately inhibit lava flow. Variations in the bulk composition of erupted lava indicate simultaneous tapping of primitive magma and more ‘normal’ summit reservoir magma(s), and the weakly-phyric character of primitive lava suggests that it may have circumvented the volcano's shallow reservoir system. Considering the 1859 eruption in the context of historic activity at Mauna Loa, we propose that periods of high magma supply may permit radial vents to tap primitive magmas with increased frequency. These rare flank eruptions may present unusual hazards because of both vent location and magma composition, thus warranting special consideration in models and mitigation efforts. Acknowledgements The authors acknowledge A. Harris and S. Rowland for helpful discussions on the operation and application of FLOWGO and J. Moore for sharing observations from his dives on the submarine portion of the 1859 lava flows. We also thank C. Thornber for generously providing unpublished data from the 1859 and 1950 eruptions of Mauna Loa. Cheers to T. Hales for graphics assistance. Finally, this manuscript

155

benefited greatly from the careful and constructive reviews of R. Helz and L. Keszthelyi. NSF awards EAR 9902851, 0207919, and 0510437 supported this project. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jvolgeores.2009.03.002. References Acocella, V., Neri, M., 2003. What makes flank eruptions? The 2001 Etna eruption and its possible triggering mechanisms. Bulletin of Volcanology 65 (7), 517–529. Alexander, W.D., 1859. Later details from the volcano on Hawaii. Pacific Commercial Advertiser. February 24, 1859, p. 2. Armstrong, J.T., 1988. Quantitative analysis of silicate and oxide minerals: comparison of Monte-Carlo, ZAF, and Phi-Rho-Z procedures. In: Newbury, D.E. (Ed.), Microbeam Analysis. San Francisco Press, San Francisco, CA, pp. 239–246. Baloga, S.M., Glaze, L.S., Crisp, J.A., Stockman, S.A., 1998. New statistics for estimating the bulk rheology of active lava flows: Puu Oo examples. Journal of Geophysical Research 103 (B3), 5133–5142. Baloga, S.M., Glaze, L.S., Peiterson, M.N., Crisp, J.A., 2001. Influence of volatile loss on thickness and density profiles of active basaltic flow lobes. Journal of Geophysical Research 106 (B7), 13,395–13,405. Barnard, W.M., 1990. Mauna Loa: A Source Book, Volume 1. W.M. Barnard, Fredonia. Barnes, H.A., Hutton, J.F., Walters, K., 1989. An Introduction to Rheology. Elsevier, Amsterdam. 210 pp. Calvari, S., Coltelli, M., Neri, M., Pompilio, M., Scribano, V., 1994. The 1991–93 Etna eruption: chronology and geological observations. Acta Vulcanologica 4 (1), 1–15. Calvari, S., Neri, M., Pinkerton, H., 2002. Effusion rate estimations during the 1999 summit eruption on Mount Etna, and growth of two distinct lava flow fields. Journal of Volcanology and Geothermal Research 119, 107–123. Cashman, K.V., Mangan, M.T., Newman, S., 1994. Surface degassing and modifications to vesicle size distributions in active basalt flows. Journal of Volcanology and Geothermal Research 61 (1), 45–68. Cashman, K.V., Thornber, C.R., Kauahikaua, J.P., 1999. Cooling and crystallization of lava in open channels, and the transition of pahoehoe lava to 'a'a. Bulletin of Volcanology 61 (5), 306–323. Cashman, K.V., Kerr, R.C., Griffiths, R.W., 2006. A laboratory model of surface crust formation and disruption on channelized lava flows. Bulletin of Volcanology 68 (8), 753–770. Ching, F.K.W., 1971. The archaeology of South Kohala and North Kona from the apuhua'a of Lalamilo to the apuhua'a of Hanamana — surface survey Kailua–Kawaihae road corridor. Hawaii State Archaeological Journal 71 (1), 260. Clague, D.A., Hagstrum, J.T., Champion, D.E., Beeson, M.H., 1999. Kilauea summit overflows: their ages and distribution in the Puna District, Hawaii. Bulletin of Volcanology 61 (6), 363–381. Clague, D.A., Moore, J.G., Dixon, J.E., Friesen, W.B., 1995. Petrology of submarine lavas from Kilauea's Puna Ridge, Hawaii. Journal of Petrology 36 (2), 299–349. Coan, T.,1860. Eruption of Mauna Loa, Sandwich Islands. American Journal of Science 29, 302. Crisp, J.A., Baloga, S.M., 1990. A model for lava flows with two thermal components. Journal of Geophysical Research 95 (B2), 1,255–1,270. Crisp, J.A., Baloga, S.M., 1994. Influence of crystallization and entrainment of cooler material on the emplacement of basaltic 'a'a lava flows. Journal of Geophysical Research 99 (B6), 11,819–11,832. Crisp, J.A., Cashman, K.V., Bonini, J.A., Hougen, S.B., Pieri, D., 1994. Crystallization history of the 1984 Mauna Loa flow. Journal of Geophysical Research 99 (B4), 7,177–7,198. Dana, J.D., 1859. Eruption of Mauna Loa, Hawaii. American Journal of Science 27, 410–415. Davis, M.G., Garcia, M.O., Wallace, P.J., 2003. Volatiles in glasses from Mauna Loa Volcano, Hawai'i: implications for magma degassing and contamination, and growth of Hawaiian volcanoes. Contributions to Mineralogy and Petrology 144, 570–591. Donaldson, C.H., 1976. An experimental investigation of olivine morphology. Mineralogy and Petrology 57 (2), 187–213. Faure, F., Trolliard, G., Nicollet, C., Montel, J.M., 2003. A developmental model of olivine morphology as a function of cooling rate and the degree of undercooling. Contributions to Mineralogy and Petrology 145 (2), 251–263. Garcia, M.O., 1996. Petrography and olivine and glass chemistry of lavas from the Hawaii Scientific Drilling project. Journal of Geophysical Research 101, 11,701–11,713. Ghiorso, M.S., Sack, R.O., 1995. Chemical mass transfer in magmatic processes IV: a revised and internal consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology 119 (2–3), 197–212. Gower, M.M., 1886. A Voyage from Boston to California and the Sandwich Islands via Cape Horn, in the Years 1858 & 1859. Green, W.L., 1887. Vestiges of the molten globe, as exhibited in the figure of the Earth. Volcanic Action and Physiography, Part II. Hawaiian Gazette, Honolulu, 337 pp. Griffiths, R.W., Fink, J.H., Cashman, K.V., 2003. Patterns of solidification in channel flows with surface cooling. Journal of Fluid Mechanics 496, 33–62. Guest, J.E., Kilburn, C.R.J., Pinkerton, H., Duncan, A., 1987. The evolution of flow fields: observations of the 1981 and 1983 eruptions of Mount Etna, Sicily. Bulletin of Volcanology 49, 527–540.

156

J.M. Riker et al. / Journal of Volcanology and Geothermal Research 183 (2009) 139–156

Hammer, J.E., Cashman, K.V., Hoblitt, R.P., Newman, S., 1999. Degassing and microlite crystallization during pre-climactic events of the 1991 eruption of Mt. Pinatubo, Philippines. Bulletin of Volcanology 60 (5), 355–380. Harris, A.J.L., Rowland, S.K., 2001. FLOWGO: a kinematic thermo-rheologic model for lava flowing in a channel. Bulletin of Volcanology 63 (1), 20–44. Haskell, R.C., 1859a. Eruption of Mauna Loa, Sandwich Islands. American Journal of Science 28, 284. Haskell, R.C., 1859b. On a visit to the recent eruption of Mauna Loa, Hawaii. American Journal of Science 28, 66–71. Haskell, R.C., 1860. Eruption of Mauna Loa, Sandwich Islands. American Journal of Science 29, 301–302. Helz, R.T., Banks, N.G., Heliker, C.C., Neal, C.A., Wolfe, E.W., 1995. Comparative geothermometry and thermal history of recent Hawaiian eruptions. Journal of Geophysical Research 100 (B9), 17,637–17,657. Helz, R.T., Heliker, C.C., Hon, K., Mangan, M.T., 2003. Thermal Efficiency of Lava Tubes in the Pu'u 'O'o-Kupaianaha Eruption. U.S. Geological Survey Professional Paper, vol. 1676, pp. 105–120. Hon, K., Kauahikaua, J.P., Denlinger, R., Mackay, K., 1994. Emplacement and inflation of pahoehoe sheet flows: observations and measurements of active lava flows on Kilauea Volcano, Hawaii. Geological Society of America Bulletin 106 (3), 351–370. Hulme, G., 1974. The interpretation of lava flow morphology. Geophysical Journal International 39 (2), 361–383. Jurado-Chichay, Z., Rowland, S.K., 1995. Channel overflows of the Pohue Bay flow, Mauna Loa, Hawai'i: examples of the contrast between surface and interior lava. Bulletin of Volcanology 57 (2), 117–126. Kaakua, J.H., 1859. No ka Pele. Ka Hae Hawai'i. November 9, 1859, p. 1. Kahookaumaha, J.A., 1859. Na hana kupaianaha a ke Akua ma Kona, Hawaii. Ka Hae Hawai'i. April 20, 1859, p. 2. Katz, M.G., Cashman, K.V., 2003. Hawaiian lava flows in the third dimension: identification and interpretation of pahoehoe and 'a'a distribution in the KP-1 and SOH-4 cores. Geochemistry Geophysics Geosystems 4 (2), 1–24. Kauahikaua, J.P., Margriter, S., Lockwood, J.P., Trusdell, F.A., 1995. Application of GIS to the estimation of lava flow hazards on Mauna Loa Volcano, Hawai'i. In: Rhodes, J.M., Lockwood, J.P. (Eds.), Mauna Loa Revealed: Structure, Composition, History, and Hazards. American Geophysical Union, Washington, D.C., pp. 315–325. Kauahikaua, J.P., Cashman, K.V., Mattox, T.N., Heliker, C.C., Thornber, C.R., 1998. Observations on basaltic lava streams in tubes from Kilauea Volcano, Hawai'i. Journal of Geophysical Research 103 (B11), 27,303–27,324. Kauahikaua, J.P., Sherrod, D.R., Cashman, K.V., Heliker, C.C., Hon, K., Mattox, T.N., Johnson, J.A., 2003. Hawaiian Lava-Flow Dynamics During the Pu'u 'O'oKupaianaha Eruption: a Tale of Two Decades. U.S. Geological Survey Professional Paper, vol. 1676, pp. 63–87. Kerr, R.C., Lister, J.R., 1991. The effects of shape on crystal settling and on the rheology of magmas. Journal of Geology 99 (3), 457–467. Kerr, R.C., Griffiths, R.W., Cashman, K.V., 2006. Formation of channelized lava flows on an unconfined slope. Journal of Geophysical Research 111, B10206. Kilburn, C.R.J., 1990. Surfaces of aa flow-fields on Mount Etna, Sicily: morphology, rheology, crystallization and scaling phenomena. In: Fink, J.H. (Ed.), Lava Flows and Domes: Emplacement Mechanisms and Hazard Implications. Springer-Verlag, Berlin, pp. 129–156. Kilburn, C.R.J., 1993. Lava crusts, aa flow lengthening, and the pahoehoe-aa transition. In: Kilburn, C.R.J., Luongo, G. (Eds.), Active Lavas: Monitoring and Modelling. University College London Press, London, pp. 263–280. Kilburn, C.R.J., 1996. Patterns and predictability in the emplacement of subaerial lava flows and flow fields. In: Scarpa, C., Tilling, R.I. (Eds.), Monitoring and Mitigation of Volcano Hazards. Springer-Verlag, Berlin, pp. 491–537. Kilburn, C.R.J., 2004. Fracturing as a quantitative indicator of lava flow dynamics. Journal of Volcanology and Geothermal Research 132 (2–3), 209–224. Lesher, C.E., Cashman, K.V., Mayfield, J.D., 1999. Kinetic controls on crystallization of Tertiary North Atlantic basalt and implications for the emplacement and cooling history of lava at site 989, southeast Greenland rifted margin. Proceedings of the Ocean Drilling Program, Scientific Results 163, 135–148. Lipman, P.W., 1980. The southwest rift zone of Mauna Loa: implications for structural evolution of Hawaiian volcanoes. American Journal of Science 280A (2), 752–776. Lipman, P.W., Banks, N.G., 1987. U.S. Geological Survey Professional Paper. Aa Flow Dynamics, Mauna Loa 1984, vol. 1350, pp. 1,527–1,567. Lipman, P.W., Banks, N.G., Rhodes, J.M., 1985. Degassing-induced crystallization of basaltic magma and effects on lava rheology. Nature 317 (6038), 604–607. Llewellin, E., Manga, M., 2005. Bubble suspension rheology and implications for conduit flow. Journal of Volcanology and Geothermal Research 143 (3), 205–217. Lockwood, J.P., Lipman, P.W., 1987. Holocene Eruptive History of Mauna Loa Volcano. U.S. Geological Survey Professional Paper, vol. 1350, pp. 509–535. Lyman, H., 1859. On the 1859 Eruption of Mauna Loa. Boston Society of Natural History Proceedings, vol. 7, pp. 38–39. 134–135. Lyons, L., 1859. The lava stream near Kawaihae! Pacific Commercial Advertiser. February 17, 1859, p. 2. Malin, M.C., 1980. Lengths of Hawaiian lava flows. Geology 8 (7), 306–308. Manga, M., Castro, J., Cashman, K.V., Loewenberg, M., 1998. Rheology of bubble-bearing magmas. Journal of Volcanology and Geothermal Research 87 (1), 15–28. Marsh, B.D., 1981. On the crystallinity, probability of occurrence, and rheology of lava and magma. Contributions to Mineralogy and Petrology 78 (1), 85–98. Montierth, C.M., 1999. Geothermometry, Crystallization, and the Pahoehoe/Aa Transition in Mauna Loa Lavas. Ph.D. Thesis, University of Oregon, 137 pp.

Montierth, C.M., Johnston, A.D., Cashman, K.V., 1995. An empirical composition-based glass geothermometer for Mauna Loa lava. In: Rhodes, J.M., Lockwood, J.P. (Eds.), Mauna Loa Revealed: Structure, Composition, History, and Hazards. American Geophysical Union, pp. 207–217. Moore, H.J., 1987. Preliminary Estimates of the Rheological Properties of 1984 Mauna Loa Lava. U.S. Geological Survey Professional Paper, vol. 1350, pp. 1,569–1,588. NaKapae, 1859. Pele Hou. Ka Hai Hawai'i. August 3, 1859, p. 1. Pal, R., 2003. Rheological behavior of bubble bearing magmas. Earth and Planetary Science Letters 207 (1), 165–179. Peterson, D.W., Swanson, D.A., 1974. Observed formation of lava tubes. Studies in Speleology 2 (6), 209–222. Peterson, D.W., Holcomb, R.T., Tilling, R.I., Christiansen, R.I., 1994. Development of lava tubes in the light of observations at Mauna Ulu, Kilauea Volcano, Hawaii. Bulletin of Volcanology 56 (5), 343–360. Pinkerton, H., Wilson, L., 1994. Factors controlling the lengths of channel-fed lava flows. Bulletin of Volcanology 56 (2), 108–120. Polacci, M., Cashman, K.V., Kauahikaua, J.P., 1999. Textural characterization of the pahoehoe-'a'a transition in Hawaiian basalt. Bulletin of Volcanology 60 (8), 595–609. Pompilio, M., Trigila, R., Zanon, V., 1998. Melting experiments on Etnan lavas: the calibration of an empirical geothermometer to estimate the eruptive temperature. Acta Vulcanologica 10 (1), 1–9. Powers, H.A., 1955. Composition and origin of basaltic magmas on the Hawaiian islands. Geochimica et Cosmochimica Acta 7 (2), 77–107. Probstein, R.F., Sengun, M.Z., Teng, T.C., 1994. Bimodal model of concentrated suspension viscosity for distributed particle sizes. Journal of Rheology 38 (4), 811–829. Rhodes, J.M., 1983. Homogeneity of lava flows: chemical data for historic Mauna Loa eruptions. Journal of Geophysical Research 88 (s2), 869–879. Rhodes, J.M., 1988. Geochemistry of the 1984 Mauna Loa eruption: implications for magma storage and supply. Journal of Geophysical Research 93 (B5), 4453–4466. Rhodes, J.M., 1995. The 1852 and 1868 Mauna Loa picrite eruptions: clues to parental magma compositions and the magmatic plumbing system. In: Rhodes, J.M., Lockwood, J.P. (Eds.), Mauna Loa Revealed: Structure, Composition, History, and Hazards. American Geophysical Union, Washington, D. C., pp. 241–262. Rhodes, J.M., Hart, S.R., 1995. Episodic trace element and isotopic variations in historical Mauna Loa lavas: implication for magma plume dynamics. In: Rhodes, J.M., Lockwood, J.P. (Eds.), Mauna Loa Revealed: Structure, Composition, History, and Hazards. American Geophysical Union, Washington, D.C., pp. 263–288. Rowland, S.K., Garbeil, H., 2000. Slopes of oceanic basalt volcanoes. In: Mouginis-Mark, P.J., Crisp, J.A., Fink, J.H. (Eds.), Remote Sensing of Active Volcanism. American Geophysical Union, Washington, D.C., pp. 223–247. Rowland, S.K., Walker, G.P.L., 1990. Pahoehoe and aa in Hawaii: volumetric flow rate controls the lava structure. Bulletin of Volcanology 52 (8), 615–628. Rowland, S.K., Garbeil, H., Harris, A.J.L., 2005. Lengths and hazards from channel-fed lava flows on Mauna Loa, Hawai'i, determined from thermal and downslope modeling with FLOWGO. Bulletin of Volcanology 67 (7), 634–647. Rust, A.C., Manga, M., 2002. Effects of bubble deformation on the viscosity of dilute suspensions. Journal of Non-Newtonian Fluid Mechanics 104 (1), 53–63. Saar, M.O., Manga, M., Cashman, K.V., Fremouw, S., 2001. Numerical models of the onset of yield strength in crystal–melt suspensions. Earth and Planetary Science Letters 87 (4), 367–379. Shaw, H.R., 1972. Viscosities of magmatic liquids: an empirical method of prediction. American Journal of Science 272 (9). Sleeper, J.H., 1859. On the 1859 eruption of Mauna Loa. Pacific Commercial Advertiser. March 10, 1859, p. 2. Soule, S.A., Cashman, K.V., 2005. The shear rate dependence of the pahoehoe-to-aa transition: analog experiments. Geology 33 (5), 361–364. Soule, S.A., Cashman, K.V., Kauahikaua, J.P., 2004. Examining flow emplacement through the surface morphology of three rapidly emplaced, solidified lava flows, Kilauea Volcano, Hawai'i. Bulletin of Volcanology 66 (1), 1–14. Stickel, J.J., Powell, R.L., 2005. Fluid mechanics and rheology of dense suspensions. Annual Review of Fluid Mechanics 37, 129–149. Stokes, J.F.G., 1991. Heiau of the Island of Hawai'i: a Historic Survey of Native Hawaiian Temple Sites. Bishop Museum Press, Honolulu. 196 pp. Tallarico, A., Dragoni, M., 2000. A three-dimensional Bingham model for channeled lava flows. Journal of Geophysical Research 105 (B11), 25,969–25,980. Tilling, R.I., Rhodes, J.M., Sparks, J.W., Lockwood, J.P., Lipman, P.W., 1987. Disruption of the Mauna Loa magma system by the 1868 Hawaiian earthquake: geochemical evidence. Science 235 (4785), 196–199. Underwood, E., 1970. Quantitative Stereology. Addison Wesley-Longman, Reading, MA. 274 pp. Walker, G.P.L., 1973. Lengths of lava flows. In: Guest, J.E., Skelhorn, R.R. (Eds.), Mount Etna and the 1971 Eruption. Royal Society of London Philosophical Transactions, London, pp. 107–118. Wildemuth, C.R., Williams, M.C., 1984. Viscosity of suspensions modeled with a sheardependent maximum packing fraction. Rheologica Acta 23 (6), 627–635. Wilkinson, J.F.G., Hensel, H.D., 1988. The petrology of some picrites from Mauna Loa and Kilauea volcanoes, Hawaii. Contributions to Mineralogy and Petrology 98 (3), 326– 345. Wright, T.L., 1971. Chemistry of Kilauea and Mauna Loa Lava in Space and Time. U.S. Geological Survey Professional Paper, vol. 735, pp. 1–39.