The 1963–1964 eruption of Agung volcano (Bali, Indonesia)
The February 1963 to January 1964 eruption of Gunung Agung, Indonesia’s largest and most devastating eruption of the twentieth century, was a multi-phase explosive and effusive event that produced both basaltic andesite tephra and andesite lava. A rather unusual eruption sequence with an early lava flow followed by two explosive phases, and the presence of two related but distinctly different magma types, is best explained by successive magma injections and mixing in the conduit or high level magma chamber. The 7.5-km-long blocky-surfaced andesite lava flow of ∼0.1 km3 volume was emplaced in the first 26 days of activity beginning on 19 February. On 17 March 1963, a major moderate intensity (∼4 × 107 kg s−1) explosive phase occurred with an ∼3.5-h-long climax. This phase produced an eruption column estimated to have reached heights of 19 to 26 km above sea level and deposited a scoria lapilli to fine ash fall unit up to ∼0.2 km3 (dense rock equivalent—DRE) in volume, with Plinian dispersal characteristics, and small but devastating scoria-and-ash flow deposits. On 16 May, a second intense 4-h-long explosive phase (2.3 × 107 kg s−1) occurred that produced an ∼20-km-high eruption column and deposited up to ∼0.1 km3 (DRE) volume of similar ash fall and pyroclastic flow deposits, the latter of which were more widespread than in the March phase. The two magma types, porphyritic basaltic andesite and andesite, are found as distinct juvenile scoria populations. This indicates magma mixing prior to the onset of the 1963 eruption, and successive injections of the more mafic magma may have modulated the pulsatory style of the eruption sequence. Even though a total of only ∼0.4 km3 (DRE volume) of lava, scoria and ash fall, and scoria-and-ash pyroclastic flow deposits were produced by the 1963 eruption, there was considerable local damage caused mainly by a combination of pyroclastic flows and lahars that formed from the flow deposits in the saturated drainages around Agung. Minor explosive activity and lahar generation by rainfall persisted into early 1964. The climactic events of 17 March and 16 May 1963 managed to inject ash and sulfur-rich gases into the tropical stratosphere.
KeywordsAgung volcano Explosive eruption Plinian deposit Scoria-and-ash flow deposit Lava flow Magma mixing
The 1963–1964 eruption of Gunung Agung stratovolcano on the island of Bali, Indonesia, is considered to be among the most important volcanic events of the twentieth century, particularly because of its possible effects on global climate (Cadle et al. 1976; Hansen et al. 1978; Self et al. 1981; Rampino and Self 1982, 1984). It was the largest magnitude and most devastating eruption of an Indonesian volcano since the 1883 eruption of Krakatau. On the island of Bali, pyroclastic flows and lahars, and to a lesser extent earthquakes, caused considerable loss of life (>1,100 deaths, Tanguy et al. 1998). We report here on the 1963–1964 eruption sequence, characteristics and volume of the products, and eruption mechanisms of this significant volcanic event.
Prior to its eruption in 1963, Agung had been dormant for 120 years. Almost nothing is known about the c. 1843 eruption, and the subsequent written history of the volcano states only that increased solfataric activity at the summit crater was observed in 1908, 1915, and 1917 (Zen and Hadikusumo 1964). The magmatic history of Agung has been discussed by Wheller (1987). Previously, the petrography and chemistry of the 1963–1964 eruption products were studied in an attempt to determine the extent of mixing in the 1963 magma batch and the nature and source of climatically significant volatiles emitted (Self and King 1996).
1963–1964 eruption sequence
The main events of the eruption have been pieced together from reports of the Volcanological Survey of Indonesia (Kusumadinata 1964a, b, c; Surjo 1964, 1965, 1981) and the paper by Zen and Hadikusumo (1964). On 18 February 18, 1963, following 2 days of locally felt tremors (Mercalli scale II) including a more significant earthquake (Mercalli scale III), minor explosive activity began in the summit vent of Agung and continued with increasing intensity through mid-March (Surjo 1981). Ejection of incandescent material and ash in pulsating, intermittent explosions, with eruption clouds reaching as high as 6 km above the summit crater (which lies at 3 km above sea level), minor pyroclastic flows of unknown source (from collapse of either vent-derived columns or the lava front), and the fall of ash- to lapilli-size particles accompanied effusion of a viscous, andesite lava flow that poured out from a pre-existing notch in the north rim of crater beginning on 19 February. The lava flowed down the north slope of the volcano before effusion from the crater was terminated on 17 March by the first of two paroxysmal explosive events. Lava front collapse avalanches were reported late on 19 February.
The first major explosive phase lasted ∼7 h on 17 March, but a 3.5-h period of climactic, high-intensity activity from 05:30 to 09:00 hours local time can be gleaned from the report of Surjo (1981). It produced a widespread ash fall (Unit 3, Fig. 4) dispersed west–northwestwards, with ash fall being reported as far away as Bandung and Jakarta in western Java, a distance of about 1,000 km from the volcano (Fig. 1). Syn-eruptive changes in the rim of the summit vent led to pyroclastic flows on the north and south sides of the volcano, which devastated the upper slopes and river canyons on the flanks of Agung. March occurs during the wet season in Bali, and the pyroclastic flow and tephra fall deposits rapidly generated hot and cold lahars by mixing with the torrential rainfalls that followed. Heavy rains on Agung began eroding the pyroclastic deposits immediately after the explosive phase of 17 March, devastating the lowlands with lahars that reached all the way to the coast (Fig. 3). The March eruption phase caused many casualties and destruction of villages, largely from the pyroclastic flows (presumed to be generated by partial column collapse) and ensuing lahars. The explosions reamed out the summit vent and crater (Zen and Hadikusumo 1964) and apparently cut off supply to the lava flow.
Explosive activity continued intermittently after the 17 March explosive phase, with small explosion clouds reported to reach 4 km above the crater rim, ash showers, and occasional small pyroclastic flows. On 11 May, explosions sent eruption clouds up to 2.5 km above the summit, then between 11 and 15 May the activity subsided. On 16 May, the second major explosive phase occurred lasting 4 to 5 h with columns reported to reach at least 10 km above the volcano summit. This phase produced a greater proportion of the total 1963 pyroclastic flow material than that of March, presumably also by column collapse, and led to much destruction in villages at the foot of the volcano. After this phase, explosive outbursts continued spasmodically from the summit vent through to 17 January 1964, generally decreasing in intensity after the end of May 1963, but occasionally punctuated by larger explosive events having a 3- to 4-km-high eruption cloud (above the vent) and yielding pyroclastic flows such as on 14 June and 22 October 1963. On 18 May, a rather strong earthquake (magnitude unknown) was reported at Rendang, and on 31 May, explosions were recorded with an eruption column reported to reach 9 km above the crater.
The 16 May paroxysm occurred during the dry season, but when the rainy season began in November 1963, lahars remobilized the pyroclastic flow deposits choking the valleys, devastating river canyons and lowlands, and causing many more deaths. Figure 3 shows the areas affected by the 1963 lava flow, pyroclastic flows, and lahars; March and May 1963 pyroclastic flows and the ensuing lahars occupied many of the same valleys, were difficult to distinguish from one another in the field, and are not discriminated on the map.
Although eyewitness reports give the eruption column heights for the two climactic phases as ∼13 km above sea level (Zen and Hadikusumo 1964), sulfuric acid aerosols and ash collected by atmospheric sampling from an aircraft over Australia at 20 km altitude from April 1963 for ∼1 year after the eruption (Mossop 1964) indicate the columns must have reached significantly higher altitudes (Rampino and Self 1982; Self and King 1996). A pictorial account of the eruption and its aftermath is given in Booth et al. (1963).
The blocky 1963 lava is typical in appearance for an eruption from a stratovolcano, but has some unusual attributes. First, besides minor explosions, it was the earliest major product of the eruption. As the initial summit activity on Agung was noticed on 18 February, lava must have rapidly filled the pre-existing summit crater and overspilled the lowest point by the night of 19 February. The single flow unit attained a length of 4 km by 24 February and reached its near-maximum length of 7 km in the 26 days before the 17 March paroxysm (Surjo 1981). Poor visibility around the volcano summit during the wet season may have precluded observations of earlier signs of filling of the crater.
Second, lava effusion and explosive events were recorded simultaneously from the central crater for almost a month, but no details are available to explain this occurrence. Two vents are a possible explanation, but the crater size and configuration were considerably changed by the major explosive events such that observations during a post-paroxysmal inspection by airplane on 6 May and 18 May were inconclusive on this point (Zen and Hadikusumo 1964). Other stratovolcanoes such as Fuego, Guatemala, have been observed to produce explosions and lava simultaneously from separate vents in the summit crater (Lyons et al. 2010).
Third, the lava flow unit lengthened after it was beheaded by the major 17 March explosive phase. An aerial inspection and a sketch made on 6 May (Zen and Hadikusumo 1964, their Fig. 2), before the second main explosive phase, indicate that the crater was not then full of lava. The blocky-surfaced lava flowed down the northern slope a distance of ∼7 km before the 17 March explosive phase; the flow was reported to have a maximum flow front height of 75 m, and 1 km in width, before 17 March, and glowing block avalanches from the lava indicate continued motion was still occurring on 3 April. Between 17 March and 3 May, the lava flow lengthened by ∼0.5 km and was observed to widen to 1.25 km and decrease in thickness by ∼15 m at the terminus (Surjo 1981). Either, after being cut off from source at the crater rim on 17 March, the lava advanced further by continued downflow of the liquid interior on the steep slopes of Agung, or the crater was rapidly refilled with lava after 17 March but had subsided again by the time of the 6 May aerial inspection (before the 16 May explosive phase). When observed in late 1979, the lava flow front was 40–45 m high.
The andesitic lava is the most compositionally evolved eruption product; phenocryst + microphenocryst content of the 1963 lava at the flow front is 24 vol% and the lava matrix glass is dacitic in composition (Self and King 1996; sample AG 49). The lava at the flow front contains numerous centimeter- to decimeter-sized angular xenolithic blocks of older lavas, presumed to be derived from the conduit and vent walls and/or perhaps deposited onto the lava in the crater by explosions.
The volume of the lava flow was given by Zen and Hadikusumo (1964) as 0.1 km3, an approximate estimate with which we agree. A bulk volume of perhaps up to 0.15 km3, based on the current flow dimensions, would convert to a dense rock equivalent (DRE) volume (including xenolithic blocks) of ∼0.1 km3 (using a melt density of 2,600 kg m−3 appropriate to the composition; Self and King 1996). Assuming a total duration of 27 days for the lava flow (to 17 March) yields a mean output rate of ∼43 m3 s−1 (or 1.1 × 105 kg s−1), ignoring the small-volume magmatic component in Fall Units 1 and 2 erupted during the same period (Fig. 4, top).
For the proximal–medial fall deposits, mapping in the field was undertaken by the authors. We have the most information on Fall Unit 3, which we interpret to have been generated during the 3.5-h-long climax of the 17 March explosive phase. Complete un-eroded sections containing the 17 March Fall Unit 3 (Fig. 4, top and bottom) were not common by the time we measured it, and we found only 36 locations at which we believe the entire Unit 3 was preserved; the stippled area on Fig. 5a shows the main area of preserved, mapable 1963 deposits, save for a few outlying remnants. The 16 May Fall Unit 5 (Fig. 4, top) was even more poorly preserved by the time we sampled it, and data collected are insufficient to produce isopach and isopleths maps. Thus, for the total dispersal of the March and May climactic fall deposits, we rely on the 1963 VSI distribution maps (Fig. 5a, c).
The accumulated fall deposit of the explosive phases in 1963 consists of several units (Fig. 4, top). The section in Fig. 4 (top) is a composite of several logged in an area 10–12 km to the west of Agung’s summit around Suter, where both major fall deposits are preserved. The earliest lithic-dominated unit we recognized is found in just two locations on the S flank of Agung and is not numbered. The poorly sorted deposit may have been generated by fallout from vent-clearing explosions near the beginning of activity (Feb. 17–19). Two other fall units (Units 1 and 2) of local dispersal found in sections near the volcano, in the western to southeastern sectors, are ascribed to activity before 17 March. Fall Unit 1 is bedded, fine- to medium-grained, lithic-rich but containing some layers rich in 1963 juvenile scoria (Fig. 4, top and bottom). It is found out to 15 to 20 km from the vent with a dispersal axis directed to the southwest and may possibly have been forming on 13–14 March (as discussed above). Fall Unit 2 is fine grained and may have been deposited in areas within 10 to 12 km to the west of the vent in a relative lull in explosive activity just before the 17 March climactic phase.
Unit 5 is thought to record the fallout from the 16 May event, which by contemporary accounts produced a similar-looking, but smaller-volume, fall deposit with a more limited dispersal (Fig. 5c). Little remained of un-reworked post-17 March fall units, including that of the 16 May major explosive phase, in the areas where we found the 1963 deposits. The dispersal axis of the 16 May fall unit is directed northwards, indicating dispersal by southerly winds, again consistent with the contemporary map (Fig. 5c). The 7- to 8-cm thickness of Fall Unit 5 (16 May) seen in the area west of the volcano (Fig. 4, top) is consistent with thicknesses shown on the contemporary fall deposit isopach map.
The axis of dispersal of the 17 March Fall Unit 3 in the area where we collected data (shaded area on Fig. 5a) is directed to the west-northwest (see also Fig. 6a, b) in agreement with the dispersal indicated by the contemporary maps. It is the coarsest and thickest scoria fall unit at all locations where we found the 1963 deposits. The elongate dispersal of Fall Unit 3 indicates a moderately high eruption column and winds from the east–southeast. Re-entrants in the isopach pattern (Fig. 6b, also seen on the contemporary whole deposit isopach map, Fig. 5a), and in the Md ф isopleths (Fig. 6d), coincide in position with the volcanic massif of Abang and the Batur caldera rim. The re-entrants may be explained by the shearing effects of low level winds around these topographic highs, which interrupted the uniform fallout of scoria lapilli and ash. The isopleth map of the mean diameter of relatively coarse, dense lithic fragments in Fall Unit 3 (Fig. 6c) does not show such a pattern, suggesting perhaps that more rapidly falling, dense clasts were not as affected by the local wind patterns.
Abundant fine material (<0 Φ; 1 mm) is absent in our Fall Unit 3 samples except at the western limit of the preserved deposit 40–50 km downwind from vent (Fig. 7, AG67), which forms the sole layer at that distance and at more distal sites. However, an abundant fines component was produced and formed the more distal, rapidly eroded ash-dominated part of the 17 March deposit (Figs. 1 and 5a). Samples from 40 to 50 km distance from source show that 97 wt% of the 17 March deposit is < 1 mm in grain size and still strongly unimodal, as are all samples of Unit 3 (Fig. 7). The contemporary map (Fig. 5a) suggests that ∼5 cm of the 17 March deposit fell in the area between 40 and 50 km west of the vent, consistent with our later field measurements of Fall Unit 3 in the same area (Figs. 5b and 6b).
The remnants of 16 May Fall Unit 5 are similar in appearance and overall grain size characteristics (Fig. 7, AG48) to Fall Unit 3, but the deposit contains fewer lithic clasts (generally <10 wt%) and has juvenile scoria of a more mafic basaltic andesite composition (Self and King 1996). May 16 sample AG48 was collected from a position 12 km NW of the vent and 6 km to the west of the unit’s dispersal axis and is more poorly sorted (1.23 σΦ) than, and has a similar grain size distribution to, the 17 March Fall Unit 3 sample AG27, collected 20 km W of the vent (Fig. 7). A later section contains further discussion of the dispersal of 17 March and 16 May fall units.
Pyroclastic flow deposits
The 1963 pyroclastic flow deposits were largely confined to stream valleys that originate near the summit of Agung, and they did not cover extensive areas. On the basis of contemporary descriptions, it appears that very little, if any, pyroclastic flows deposited directly into the sea. Instead, much of the original volume was remobilized into lahars during the ensuing rainy seasons (∼40 % after the first; Kusumadinata 1964c), which transported material into the sea off the north and south coasts (Fig. 3).
While some dense fragmental flows may have been caused by spalling of blocks from the lava flow on the steep, 20° to 30°, north slope of the Agung cone, as described at the beginning of lava flowage by Surjo (1981), most pyroclastic flows that filled the gullies and valleys around the volcano with deposits (Fig. 3) were probably due to eruption column collapse during the 17 March and 16 May explosive phases. This is supported by the observation that they are matrix-supported scoria-and-ash flow deposits, containing juvenile “cauliflower”-shaped bombs with a range of sizes up to several meters and variable vesicularity, together with dense lithic blocks. The 16 May event produced the greater portion of pyroclastic flow material (Surjo 1965), perhaps due to a less intense eruption rate (see later section) and more prevalent eruption column collapse. Possible co-ignimbrite ash clouds were photographed while the 16 May explosive phase was waning (Fig. 2c).
During field work, it was difficult to positively identify the major pyroclastic flow deposits of the 17 March eruption because the same drainages were occupied by 16 May flow deposits and also because of extensive erosion of the valley-filling early flows during the close of the first wet season. Flow deposits of 16 May are black to dark gray, poorly sorted, and were observed to form non-size graded units from ∼1 m to in excess of 3 m thick. Carbonized vegetation and occasional gas escape pipes in the matrix around the vegetation were seen. Cut rectangular blocks from basaltic andesite bombs (some having “cauliflower”-textured crusts) with the most common degree of vesicularity have a mean density of 1,600 kg m−3 (average of 3); typical densities are in the 1,500–1,620-kg-m−3 range. These blocks appear quite dense and poorly vesicular. The volume of pyroclastic flow deposits is difficult to determine because of erosion and generation of lahars. A bulk minimum volume is estimated at ∼0.1 km3.
Dispersal of main fall units and eruptive volume estimates
17 March and 16 May fall unit dispersal and volume
Pyroclastic fall deposits produced during the climatic explosive phases of 17 March and 16 May contain the greatest amount of new magma erupted from Agung in 1963. Areas covered by the two fall layers can be estimated from isopach maps made by VSI personnel immediately after the eruption (Fig. 5a, c) and from Fig. 4 in Zen and Hadikusumo (1964). While the areas and thicknesses on Bali are moderately well constrained, the extent of the more distal ash fall is not. This sub-section discusses the dispersal and volume estimates of the 17 March and 16 May fall units, including the uncertainties.
An inset on the VSI map of the 17 March deposit (Fall Unit 3, Fig. 4) shows the “trace” isopach (Fig. 5a) to enclose the city of Surabaya on Java and Madura Island, just off Java (short-dashed ellipse, Fig. 1). This isopach lies a maximum distance of 370 km from source along the dispersal axis; the 1-cm isopach is shown to just reach the east coast of Java NW of Bali. While Zen and Hadikusumo (1964) and Surjo (1981) both mention far-distal ash fall over Jakarta and Bandung in western Java (∼1,000 km from source), only Zen and Hadikusumo show a dispersal distribution for the far-distal ash fall (dashed outer ellipse, Fig. 1) and the information on which it is based is not discussed by them. Uncertainties for the 17 March fallout distribution include the thickness value of the “trace” isopach and whether there was a coherent layer of fine ash across Java to beyond Jakarta, or just occasional depositional nodes.
Within 50 km of source, the 1963 maps generally agree with our independent measurements made in the field almost 20 years later, allowing for thinning due to grain settling and compaction relative to the freshly fallen deposits measured by VSI personnel. While the contemporary maps are labeled as recording the thicknesses of deposits formed only on the days of paroxysmal eruptions (17 March and 16 May), we see a better fit for the 17 March deposit map with our map of Units 1, 2, and 3 combined (Fig. 5b); thus, the VSI map may incorporate deposits from 13 to 17 March. However, the volume of Fall Units 1 and 2 is very small and would affect only the second decimal place in the estimates discussed below.
Areas covered by ash fall were obtained by tracing or drawing the isopachs onto enlarged, scaled maps of Bali and Java because the original VSI maps were not on a planimetric base. Volumes were obtained from the resulting isopach maps following the methods suggested by Pyle (1989) and Fierstein and Nathenson (1992). More recent work has shown that the volumes of fall deposits can be more accurately estimated in cases of near-perfect preservation or data sets (Bonadonna and Houghton 2005) and that a common source of uncertainty is the lack of a detailed distal ash record, which is certainly the case for the 1963 deposits.
Value of trace isopach (cm)
Value of far-distal isopach (cm)
Extra isopach (cm)
Bulk vol (km3)
Bulk vola <1 cm (km3)
1: VSI fresh ash map
2: Mapped FU 1–3
3: Mapped FU 3
Case 1 (on Fig. 8) is the VSI isopach map and new assumed values (0.1 and 0.01 cm) for the two distal isopachs, which yields a bulk volume of 0.60 km3. Varying the value of the “trace” isopach to 3 mm increases the volume to 0.8 km3, demonstrating the influence of a single value on the result obtained. Case 2 is based on our reconstructed map of Fall Units 1–3 on Bali (Fig. 5b) that agrees quite well with the VSI dispersal map (Fig. 5a). We plot a 2-cm isopach interpolated from two data points, the VSI 1-cm isopach (assumed to be 0.5 cm to account for post-depositional compaction), the “trace” isopach (assumed to be 0.1 cm), and no far-distal isopach; this case yields 0.35 km3 extrapolating to 0.001 cm. Case 3 is the same as case 2 except the proximal values are based upon our field data from Fall Unit 3 only (Fig. 6b), which leads to a lower slope of the plot and a slightly larger bulk volume of 0.36 km3. Both cases 2 and 3 yield much small proportions of the volume in the part of the deposit <1 cm, compared with case 1 (Table 1).
Dispersal of the 16 May Fall Unit 5 is less well constrained than that of Fall Unit 3 as the deposit fell at sea beyond ∼20 km N of the vent (Fig. 5c) and was largely eroded away before we examined it. The VSI map shows the trace thickness limit (assumed to be 0.1 cm) extending much beyond the 1 cm isopach and reaching north of the island of Sempanjang, ∼175 km offshore (labeled Sj in Fig. 1). The curve plotted in Fig. 8 (for data used, see Appendix) is based on extrapolation to infinity, giving a bulk volume estimate of 0.34 km3 with 0.03 km3 beyond the 1-cm isopach. Thus, the original volume of the 16 May deposit is estimated to be a little smaller than that of 17 March.
A reasonable estimate for the bulk volumes of Fall Units 3 (March) and 5 (May), given the restricted data available, is ∼0.36 (rounded to 0.4) km3 and 0.34 (rounded to 0.3) km3, respectively. The volume estimated in case 1 from the contemporary VSI map is subject to considerable uncertainty due to the unconstrained distal thicknesses, with almost half the calculated volume being outside the 1-cm isopach (Table 1). Further uncertainty in the density of the freshly fallen deposit also compounds the estimate of dense rock equivalent volumes and this larger value is not considered further.
Volume estimates for the 1963 eruption products
Summary of bulk and DRE volumes (cubic kilometer) of the 1963 eruption products from Agung
Volume lava, bulk/DRE
Volume pfd, bulk/DRE
Volume 17 March, fall bulk/DRE
Volume 16 May, fall bulk/DRE
Total volume, DRE
Volume magmaa, DRE
The total volume of pyroclastic flow material deposited during the 1963 eruption is estimated at ∼0.11 km3, or 0.05 km3 DRE (Table 2), in broad agreement with the (assumed) bulk volume of 0.1 km3 given by Kusumadinata (1964a; see also Surjo 1981). Thus, the total ejecta that was explosively erupted in 1963 was at least ∼0.28 (17 March and 16 May fall deposits) plus the 0.05 (pyroclastic flow deposits), or ∼0.33 km3 DRE. Of this, based on lithic clast contents in Fall Unit 3 samples that vary from 23 wt% proximally to 12 wt% distally (see Fig. 6d for sample distribution), and also applying this lithic content to the pyroclastic flow deposits, an average of about 17 wt% non-juvenile content is deducted from the >1-cm thick part of the fall volumes and the pyroclastic flow deposits. This reduces the explosively erupted magmatic DRE volume to ∼0.30 km3. The lava flow is about 0.15 km3 and contributed ∼0.10 km3 DRE of magma to the total erupted volume.
In summary, the explosively ejected DRE volume was ∼0.33 km3 and, including the 0.1 km3 DRE lava flow volume, gives a total of about 0.43, or ∼0.4 km3 DRE for the whole eruption. Self and King (1996) estimated the total volume of eruption products to be ∼0.95 km3 DRE, which, as we show above, is probably an upper bound based on certain assumptions about the 17 March fall deposit. This new estimate suggests a lower total eruptive volume and reduces the estimated magmatic component that contributed to atmospheric release of climatically relevant gases by about 50 % from that previously suggested.
Mechanisms of the 1963–1964 Agung eruption
We here consider the magmatic and eruptive mechanisms that contributed to the 1963–1964 event. Two magma types were erupted, porphyritic basaltic andesite and andesite in composition, represented in the pyroclastic deposits as distinct juvenile scoria populations. Details are given in Self and King (1996) and only a brief summary is given here. The early formed lava flow, most juvenile clasts in Fall Units 1 and 2 and much of Fall Unit 3 are andesite. Sparse clasts of basaltic andesite magma are found in Fall Unit 1 and the lower part of Unit 3, and this type dominates in the upper part of Fall Unit 3. Fall Units 4 and 5 also contain mainly basaltic andesite magmatic clasts and this type dominates the scoria bomb component in the sampled pyroclastic flow deposits (inferred to be from the 16 May event). In addition to the juvenile scoria, identified by its fresh, vesicular nature in larger clasts, various types of andesite and basaltic andesite occur as lithic fragments in all units.
Petrographic and compositional evidence summarized in Self and King (1996) strongly indicates magma mixing prior to and during the 1963 eruption. This includes whole-rock chemical trends toward generally increasing MgO and compatible trace element contents and decreasing SiO2 and incompatible trace element contents through the eruption sequence, reverse and complex compositional zoning of mineral phases, disequilibrium mineral assemblages, slight dissolution of some crystal rims, sieve-textured plagioclase phenocrysts, and augite rims on reversely zoned orthopyroxene grains. Magma mixing is common in calc–alkaline systems, and the 1963 lava is thought to represent the more evolved end-member, while the most mafic of the 16 May basaltic andesite scoria and pyroclastic flow clasts represent the least evolved compositions erupted. Before the eruption, and possibly as a trigger, basaltic magma was introduced into a magma chamber primarily containing andesitic magma, some of which was extruded to form the early explosive phases and accompanying lava. Mixing of basalt and andesite produced basaltic andesite magma and the products were progressively released during the eruption. This may also explain the mixture of normally zoned, reversely zoned, and complexly zoned crystals in the 1963 tephra.
The chronologic sequence of the Agung eruption and the presence of two magma types lead us to propose the following sequence and mechanism. Due to injection of basaltic magma into a high level magma body of andesite, the uppermost andesite magma filled the crater around 15–18 February 1963, and overflowed, forming the lava flow. This magma movement was accompanied by mild to moderate spasmodic explosive activity. On 16–17 March, a second pulse of more mafic magma intruded the conduit, and possibly the lava lake in the crater, physically mixed with magma already in the conduit, and caused violent exsolution of volatiles. A sustained explosive eruption resulted with a more-or-less maintained column, depositing Fall Unit 3. Scoria-and-ash pyroclastic flows were generated at the same time by column collapse. The magma pulse was by then exhausted, and the lithic-rich fall layers grouped as Unit 4 probably record the waning explosive phases of the 17 March eruption.
After 2 months of intermittent activity, the magma in the conduit and possibly in the crater, capped by a thick pile of debris and perhaps a cooled carapace, had developed sufficient volatile pressure to break the cap, and this depressurization led to a second violent volatile exsolution event, forming a slightly smaller fall deposit on 16 May. There may have been a second pulse of mafic magma before this event. The eruption column during the May 16 event reached high altitudes and pyroclastic flow deposits again formed by column collapse. Activity then subsided gradually during the ensuing months until January 1964.
Dispersal characteristics of 1963 fall units
Regarding the dispersal of the main fall units of March and May 1963, the slopes of the curves on Fig. 8, by comparison with plots of Plinian and Etnean sub-Plinian basaltic fall deposits (gray-shaded fields, after Houghton and Gonnermann 2008), have thinning characteristics of low-dispersal Plinian deposits. The Agung fall deposits also share similar dispersal characteristics with prehistoric Plinian andesitic fall deposits from Ruapehu volcano, New Zealand (Fig. 8b), described in a recent study (Pardo et al. 2012). Likewise, the 1974 basaltic andesite fall deposit from Fuego volcano (Rose et al. 2008) shows similar thinning behavior (Fig. 8b).
The missing (eroded) part of the 17 March fall unit lies mainly beyond the 5-cm isopach (Fig. 6b) and we here attempt to assess its character. At that distance from the vent (40–50 km), the unit is mainly composed of sub-1 mm ash (97 wt% in sample AG67, Fig. 7). Further, we estimated above that a considerable proportion of the erupted volume was deposited beyond ∼5 cm thickness (57–65 %, depending on the slope model used). This indicates that a large portion of the mass of the 17 March unit, composing the medial to far-distal ash fall, was composed of fine ash. While the lack of information on the distal deposit precludes us from estimating a total grain size distribution, such high proportions of fine ash are typical for total grain size distributions of similar deposits from other basaltic andesite stratovolcanoes such as Fuego, and other sub-Plinian to Plinian fall units (Rose et al. 2008; Durant and Rose 2009). The overall grain size character of the 16 May fall unit was probably similar to that of the 17 March deposit.
Estimates of eruption rate and column height
For the two 1963 climactic explosive phases, 17 March and 16 May, eyewitness reports give the column heights as only 10 to 13 km above sea level (Zen and Hadikusumo 1964), yet evidence from atmospheric aerosols indicates the columns must have reached significantly higher elevations. An increase in stratospheric aerosol optical depth (a measure of opacity to sunlight) following the eruption (see Rampino and Self 1984; Self and King 1996) further suggests that the 15–17-km-high tropopause over Bali was exceeded by the Agung eruption columns.
Deposit and eruption column characteristics for 17 March and 16 May Agung fall deposits and comparable deposits
Mass eruption rate (kg s−1)
Column heighta,b (km)
Column heightc (km)
Column heightd (km)
4 × 107
2 × 107
7 × 107–6 × 108
3 × 106
Various models (Wilson et al. 1978; Wilson and Walker 1987) are used to estimate a column height above vent for the 17 March explosive phase, summarized in Table 3. Heights between 15 and 20 km are indicated for a volumetric eruption rate of ∼2 × 104 m3 s−1, depending on efficiency of heat use in the plume and exsolved magmatic water contents between 1 and 5 wt%. Based on relationships using the isopleth map of average maximum diameter of lithics for Fall Unit 3 (Fig. 6c), another model (Carey and Sparks 1986; their Figs. 7, 15, and 16) predicts a column height between about 19 and 25 km, assuming no wind, and between about 17 and 23 km assuming a maximum crosswind at the tropopause of 30 m s−1. Their plots of crosswind range vs. maximum downwind range for varying diameter lithic particles predict a 17 March column height of about 19 to 23 km and a wind speed of 25 to 30 m s−1. We estimate the 16 May column height to have been between 13 and 17 km above vent on the same basis as above; the lack of clast-size data prevents a more detailed assessment.
These models predict a column height for the first paroxysmal phase in the range of 16 to 23 km above the Agung vent, or a total of 19 to 26 km above sea level, and a slightly lower height (16–20 km) for the second phase. Column heights within this range would be capable of injecting ash and gas into the stratosphere at the low latitude of Bali (8° S), in agreement with atmospheric evidence. Moreover, the tropical climate of Bali may favor column heights toward the higher end of these ranges due to increased plume buoyancy caused by heat released by condensation of entrained atmospheric water vapor (Woods 1993). By contrast, the 15 June 1991, ∼3.5-h-long, climactic eruption phase of Mount Pinatubo had a mass eruption rate of almost 109 kg s−1 and an eruption column ∼35 km high (e.g., Self et al. 2004).
The twin climaxes of the 1963–1964 Agung eruption were two moderate intensity (>107 kg s−1 magma output rate) explosive phases of modest magnitude (∼0.2 to 0.1 km3 DRE volume) yielding scoria and ash fall deposits with Plinian to sub-Plinian dispersal characteristics accompanied by generation of pyroclastic flows. The total volume of basaltic andesite material erupted is estimated to be <0.5 km3 (0.1 km 3 lava; <0.4 km3 pyroclastic fall deposits; 0.05 km3 pyroclastic flow deposits—all dense rock equivalent volume). The explosive eruption columns on 17 March and 16 May reached >20 km, penetrating the tropopause and causing stratospheric injection of ash and sulfur gases. The volume of magma erupted explosively and that contributed to degassing was ∼0.3 km3, less (by ∼50 %) than estimated earlier by Self and King (1996). This implies that a greater mass of degassed volatiles was derived from un-erupted magma and/or an undetermined sequestered magmatic gas phase.
This information is of great importance to the understanding of the effects of volcanoes on climate. The increase in stratospheric aerosol optical depth especially in the Southern Hemisphere, following the eruptions suggests that 10 to 12 Mt of H2SO4 aerosols were produced by the Agung eruption (Rampino and Self 1982). Frequent, high-intensity sub-Plinian to Plinian eruptions from the world’s many active basaltic/andesitic composite volcanoes such as Agung 1963–1964 are probably the most common cause of perturbations of the stratospheric aerosol layer.
We thank the Indonesian Volcanological Survey, Bandung, West Java, for their hospitality and assistance. K. Kusumadinata provided information and unpublished data; M. Samud and M. Santoso provided assistance in the field. The Indonesian Institute of Science (LIPI) kindly granted permission to work on Bali in 1979. Field work was supported by NASA grant NSG5145. Reviews by J. Fierstein and J. L. Macias, and comments by Associate Editor J. Gardner, considerably improved an earlier version of the manuscript.
- Booth PW, Matthew SW, Sisson RE (1963) Bali’s sacred mountain blows its top. Nat Geograph 124:436–458Google Scholar
- Kusumadinata K (1964a) The eruption of the Agung volcano in Bali in 1963. Bull Geol Surv Indonesia 1.1 (Notes):12–15Google Scholar
- Kusumadinata K (1964b) Renewed activity of the Agung volcano in January 1964. Bull Geol Surv Indonesia 1(2):38Google Scholar
- Kusumadinata K (1964c) Lahars of the Agung volcano as a secondary destructive element. Bull Geol Surv Indonesia 1(2)Google Scholar
- Pardo N, Cronin S, Palmer A, Proctor J, Smith I (2012) Andesitic Plinian eruptions at Mt. Ruapehu: quantifying the uppermost limits of eruptive parameters. Bull Volcanol. doi:10.1007/s00445-012-0588-y
- Surjo I (1964) Lahar of the Agung volcano after the eruption of 1963. Bull Geol Surv Indonesia 1(2):39–43Google Scholar
- Surjo I (1965) Casualties of the latest activity of the Agung volcano. Bull Geol Surv Indonesia 2(1):22–26Google Scholar
- Surjo I (1981) Report on the volcanic activity in Indonesia during the period 1961–1963. Bull Volcanol Surv Indonesia No 104:58–94Google Scholar
- Wheller GE (1987) Petrogenetic studies of Indonesian volcanism. PhD thesis, University of Tasmania, 370 ppGoogle Scholar