Bulletin of Volcanology

, Volume 72, Issue 8, pp 961–970 | Cite as

The coalescence and organization of lahars at Semeru volcano, Indonesia

  • E. E. Doyle
  • S. J. Cronin
  • S. E. Cole
  • J.-C. Thouret
Research Article


We present multi-parameter geophysical measurements of rainfall-induced lahars at Semeru Volcano, East Java, using two observation sites 510 m apart, 11.5 km from the summit. Our study site in the Curah Lengkong channel is composed of a 30-m wide box-valley, with a base of gravel and lava bedrock, representing an ideal geometry for high density measurements of active lahars. Instrumentation included pore-pressure sensors (stage), a broad-band seismograph (arrival times, vibrational energy, and turbulence), video footage, and direct bucket sampling. A total of 8 rainfall-induced lahars were recorded, with durations of 1–3 h, heights 0.5–2 m, and peak velocities 3–6 m/s. Flow types ranged from dilute to dense hyperconcentrated flows. These recorded flows were commonly composed of partly coalesced, discrete and unsteady gravity current packets, represented by multiple peaks within each lahar. These packets most likely originate from multiple lahar sources, and can be traced between instrument sites. Those with the highest concentrations and greatest wetted areas were often located mid-lahar at our measured reach, accelerating towards the flow front. As these lahars travel downstream, the individual packets thus coalesce and the flow develops a more organised structure. Observations of different degrees of coalescence between these discrete flow packets illustrate that a single mature debris flow may have formed from multiple dynamically independent lahars, each with different origins.


Lahar Hyperconcentrated flow Debris flow Velocity Sediment concentration Seismometer Density 


Lahars are fast moving mixtures of sediment and water on volcanoes. They are an extremely hazardous phenomenon in many volcanic regions, during and following eruptions, having caused an estimated 30,734 fatalities in the 20th Century alone (Witham 2005). Trigger mechanisms include crater-lake outbreaks, eruption-induced snow melting, and rainfall induced remobilization of sediment. Despite their frequency, lahars remain poorly understood and their impacts very difficult to predict. They often begin as watery floods that entrain sediment to develop a complex physical behaviour transforming between hyperconcentrated flows and debris flows (Cronin et al. 1997; Pierson 2005). Hyperconcentrated flows contain 20–60 vol.% particles (Beverage and Culbertson 1964), with a concentrated bedload region and an upper dilute suspension region, displaying aspects of non-Newtonian behavior (Pierson 2005). Debris flows are largely unstratified mixtures of water and >60 vol.% sediment, displaying a high yield stress (Coussot and Meunier 1996; Iverson 1997). These flows often appear as waves or surges with steep high concentration fronts and dilute tails.

Due to their sudden onset and hazardous nature, scientific observations of active lahars are scarce (e.g. Cronin et al. 1999; Manville and Cronin 2007). Thus, most numerical models and physical understandings are based on purely theoretical functions to describe complex processes occurring within these flows. Simplified, continuum and two-phase models of lahars (e.g. Macedonio and Pareschi 1992; Schilling 1998; O’Brien 1999; Iverson and Denlinger 2001) are important for hazard assessment, as they allow for large scale simulations over natural terrains with existing computational power. However, they are commonly limited by an inability to capture the complexities of real lahars. These models commonly assume one dominant rheology, constant concentration, and rarely address changes in volume or sediment concentration over the propagation distance.

Obtaining detailed scientific observations of these flows in motion not only furthers our understanding of their evolution downstream, but also helps develop and qualify numerical descriptions of these processes for use in hazard models. Empirical investigations focus on flume-experiments (e.g. Iverson et al. 1997; Major 1997) and rain triggered mass-flows (e.g. Lavigne et al 2003; Arattano and Moia 1999). In this paper, we present results from multi-parameter data of rainfall-induced lahars recorded on Semeru Volcano, East Java, and illustrate that a multi-peaked lahar observed at one site may actually be composed of a number of partially coalesced individual lahars generated from multiple spatial and temporal sources.

In the Curah Lengkong channel on Semeru Volcano, we use two observation sites 510 m apart, 11.5 km from the summit (Fig. 1). The use of two closely located sites provides an ideal opportunity for frequent, detailed, measurements of active lahars in a natural open channel, providing vital data to better characterize and understand the processes and downstream evolution of lahars. This field campaign also provided an opportunity to test and develop robust repeatable techniques for gauging and monitoring lahars, assessing their size, sediment concentration and turbulence. Semeru Volcano is an ideal site for this study, as rain-triggered lahars are produced on a daily to weekly basis during the rainy season. During a 3 week period in February and March 2008, we recorded a total of 8 lahars, described further in “Field Survey 2008”. We illustrate in “Event 05/03/08” how spatial variations in concentration influence the behaviour of these flows, identifying in “Estimating particle concentrations” approximate tools to estimate particle concentration from seismic signals. The effect of particle concentration on flow evolution, and the self-organisation of these lahars downstream, is then discussed in “Packet migration”.
Fig. 1

Location map and cross sectional profiles of measurement sites. There are many minor river tributaries on the flanks of the volcano (grey region), and only the largest are indicated (from the boundary of the grey region and beyond)

Geophysical recordings of lahars

Geophysical observations of lahars have utilized a wide range of instruments from ultrasonic and radar sensors to measure changes in flow surface height, through to load cells to measure basal forces, impact sensors to measure sediment transport, and electromagnetic Doppler speedometers for velocities (e.g. Lavigne et al. 2000; Marchi et al. 2002; Hürlimann et al. 2003; Rickenmann and McArdel 2007, and a full summary in Arattano and Marchi 2008). The most promising of these geophysical instruments for lahar studies are seismometers. Seismic signals are generated by the frictional interaction with channel walls, turbulent splashing, wave-breaking and particle collisions (Marcial et al. 1996; Cole et al. 2009). The magnitude of these induced ground vibrations has been correlated to flow depth, grain size (Arattano and Moia 1999), and flow discharge (Suwa et al. 2000). Physical changes in the flow and grain size distribution produce different (but non-unique) frequency responses, providing a potentially useful non intrusive tool to assess changes in flow behaviour, such as wetted area, discharge, particulate concentration, and the degree of flow turbulence.

To monitor lahars at Merapi volcano, Lavigne et al. (2000) utilized a frequency band of 8.75 to 9.25 Hz. Zobin et al. (2009) observe that the frequency content of signals induced by lahars are generally higher than those produced by pyroclastic flows, with peak frequencies of 6–8 Hz and 3–4 Hz respectively, and attribute this higher frequency to a stony or turbulent-muddy lower layer within which frequent collisions occur. However, for deployments where the seismometers are installed directly into the channel banks, the seismic signals are commonly spread across a wider range from 5–100 Hz for both hyperconcentrated and debris flows (Marcial et al. 1996; Lavigne et al. 2000; Huang et al. 2004; Cole et al. 2009).

Peak ground vibration frequencies of <50 Hz are often associated with debris flow fronts, shifting to 50–100 Hz generated by their tails (Okuda et al. 1979; Suwa et al. 2000; Huang et al. 2007). For laboratory flows, sliding frictional bed loads have been observed to induce lower frequencies (20–80 Hz) than those from particle collisions (10–500 Hz; Huang et al. 2004). Real-world laminar, sliding, snow slurry lahars at Ruapehu volcano, N.Z., have been observed by Cole et al. (2009) to be dominated by frequencies of 5–20 Hz, while turbulent hyperconcentrated flows had additional dominant high-energy vibrations above 30 Hz.

Field survey 2008

Installed instruments and data collection

In the measured reach the Lengkong river channel is an approximately 30 m wide box-valley with a gravel and lava bedrock floor. This field location has previously been utilised by Lavigne et al. (2003), and Lavigne and Suwa (2004). In our study, two instrument sites were located c. 510 m apart, with an average slope of approximately 3° between. The upstream ‘lava’ site is characterized by a u-shaped channel with a bottom width of c. 10 m, widening to 25 m at 3 m height, illustrated in Fig. 1. The downstream ‘sabo’ site has a base width of c. 20 m, broadening to c. 35 m at 2.5 m height. Instrumentation at both sites included:
  • Pore pressure sensors (Hobo U20 and Solinst) buried mid-channel in the hard-rock beds, and recording at 10 samples per second (sps) at the upstream site and 2 sps downstream. When corrected for barometric measurements of atmospheric pressure, the lahar’s hydrostatic pressure can be converted to its depth, assuming a correction factor of 1 kpa = 0.1022 m of water. The calculated stages are verified with video records, and this method has been found to agree to within 5% of radar stage records for the Ruapehu 2007 crater-lake outbreak lahar, where differences arise due to wave breaking and splashing causing radar scattering.

  • A 3 component Guralp CMG-6TD broadband seismometer installed 10 m downstream of the upstream ‘lava’ site, where it was buried into the left bank (looking downstream) with the North axis aligned parallel to the river bank. This recorded ground vibrations up to the Nyquist frequency of 62.5 Hz. These signals can be used to identify arrival times, and infer the energy (and thus discharge) of the flow, its turbulence and particle concentration (see “Estimating particle concentrations”).

  • When safely possible, direct suspended load sampling was conducted at approximately 10 to 15 min intervals at the downstream ‘sabo’ site, by regularly dipping a 10 L bucket into the lahars. This provided samples for estimates of particle concentration, grain size distribution and rheological properties (Lavigne and Suwa 2004). Rheometric tests on this collected material are discussed further by Dumaisnil et al. (2010).

  • Fixed 25 fps video cameras were mounted on tripods on the true left (looking downstream) bank of both instrument sites. This footage is used to verify arrival times, stage height, velocities, and to provide qualitative information about flow turbulence and particle concentrations.

A Real Time Kinetic-GPS survey was also conducted (with <3 cm precision) during instrument installation, providing cross-sectional profiles of the channel at each station. These are assumed to be a good approximation throughout the full recording period, allowing for the calculation of the flow cross sectional area (wetted area) from the pore pressure stage records.

General characteristics of observed flows

Average annual rainfalls of 2,000 mm to 3,250 mm have been recorded from 1976 to 2000 in the Curah Lengkong channel, and most debris flows and hyperconcentrated flows are produced during the rainy season from October through April (see summary in Lavigne and Suwa 2004). These lahars are commonly triggered by two types of rainfall, either intense stationary rainfall exceeding 25 mm/h, or low intensity migratory rainfall over several hours. During the 3 week installation period of February-March 2008, a total of eight rain-induced lahars were recorded. The flows had durations of 1–3 h, and occurred 10–15 min after the onset of heavy rainfall in the summit region.

Origin sites of the lahars are believed to be between 8 and 11.5 km from our observation reach (Fig. 1). Bulk properties ranged from hyperconcentrated streamflows up to rare coarse, but non-cohesive, debris flows. Average grain size distributions indicate 0–12.9% gravel, 71.5–90% sand and 10–16% silt and clay (Dumaisnil et al. 2010), at an average concentration of 23 ± 18 vol.%, with a maximum of 67 vol.% and minimum of 6 vol.%. Flow depths averaged between 0.5–2 m, with peak travel velocities of 3–6 m/s. Maximum observed discharges of 25–250 m3/s compare to discharges of 50–500 m3/s recorded at the downstream ‘sabo’ site (e.g. Lavigne et al. 2003).

The lahars we evaluated were usually characterized by a rapidly rising emergent onset. Estimates of the time-dependent local Froude number at each site indicate rapidly varied, unsteady, subcritical-turbulent flow for most of the lahar duration, tending to critical and supercritical conditions during the flow peaks. Laminar flow, and a decrease in surface wave breaking, is observed during the highest concentration phases. Typically, several individual event arrivals can be identified in the stage and wetted area records (Fig. 2a and b), video observations and seismic signals. Pulses or surges of this nature have previously been observed for lahars in this channel (Lavigne and Suwa 2004), and at other locations (e.g. Arattano and Moia 1999; Marchi et al. 2002; Zanuttigh and Lamberti 2007). For the lahars we have recorded, these individual arrivals can be identified and traced between the observation sites and are herein referred to as ‘packets’ (“Event 05/03/08”, and Doyle et al. 2009). These lahar packets commonly have durations of 400–2,000 s and thus cannot be roll wave phenomena, which commonly have periods ≤100 s (Arattano and Moia 1999; Takahashi 2007; Zanuttigh and Lamberti 2007). In addition, the low channel gradient (approximately 3°) and subcritical flow conditions would not be suitable for the formation of roll waves.
Fig. 2

Example of the wetted area calculated from pore-pressure stage records of 7/3/08. Dotted and solid vertical lines indicate packet arrival times at the upstream (a) and downstream (b) sites respectively

Long discrete surges within lahars have been attributed to spatially and temporally distributed lahar sources; damming, ponding or surging of the flow; or by secondary entrainment of bed material via the dilute tail sections of initial surges (Iverson 1997; Arattano and Moia 1999; Marchi et al. 2002; Takahashi 2007; Zanuttigh and Lamberti 2007). At our field location, Lavigne and Suwa (2004) have identified that there can be distinct changes in rainfall intensity prior to and during a lahar event, with different types of rainfall occurring, from stationary to migratory. We interpret each of the packets we observe to be dynamically independent lahars, originating from different sources or tributaries during a single meteorological event (Fig. 1). These coalesce in a single channel, leading to the observation of a multi-peaked lahar at our observation sites.

Event 05/03/08

We now describe in detail the event of 5 March 2008, which has the fullest available data set and illustrates a typical progression from dilute through to hyperconcentrated flow. The packets are not characterised by a blunt front, as has been observed at this location for high concentration debris flows (Lavigne et al. 2003). Rather, the onset of each packet is commonly identified by an emergent and rapid increase in stage and wetted area (Fig. 3a and d). In addition, arrivals can be interpreted from video and visual observations, and from sharp increases in the seismic signal (Fig. 3b), and the approximate spectral energy of 512-point Fast Fourier Transform (FFT) spectra, where the latter is calculated throughout the lahar with 50% overlapping windows (Fig. 3c). A typical noise spectrum was calculated by averaging the 512 point amplitude frequency spectra calculated from four one hour signals: before and after the lahar, and from the day before and after. This indicates low frequency volcanic noise of <5 Hz, which has been filtered from the signal prior to analysis.
Fig. 3

Upstream ‘lava’ site data on 5/3/08: a 20 s smoothed wetted area, b cross-channel component of seismic ground motion, filtered <5 Hz to remove volcanic noise, c approximate spectral energy calculated from integrations of 512 point FFT velocity amplitude spectra, d downstream ‘sabo’ site data: 20 s smoothed wetted area (line) and sampled volume concentrations (circles). Arrows in (a) illustrate the equivalent sample times at the ‘lava’ site. Vertical lines indicate packet arrival times at both sites. e Averaged 512 point FFT velocity amplitude spectra for each packet in (b)

Bucket samples (“Installed instruments and data collection”), illustrate that the sediment maxima lag the lahar front by approximately 33 min (Fig. 3d). There is also a correlation between the sampled particle concentration and the front velocity of each packet. This velocity Vf is calculated as an average propagation velocity between sites, based on the travel distance and time. Packet 1 has a sampled particulate concentration φ downstream of ∼26 vol.% and travels at Vf = 1.5 ± 0.1 m/s between sites, packet 2 has φ ≈ 48 vol.% and Vf = 2.9 ± 0.2 m/s, packet 3 has φ ≈ 60 vol.% and Vf = 4.0 ± 0.3 m/s, and packet 4 has φ ≈ 40% and Vf = 1.8 ± 0.1 m/s. The higher concentration, faster moving packets are accelerating with respect to the slower moving flow front (Doyle et al. 2009). In addition, the wetted area of the higher concentration packets 2 and 3 increases downstream, from peaks of 6.1 and 6.5 m2 to 10.3 and 8.9 m2, respectively (Fig. 3a and d).

A correlation between particle concentration and flow behaviour is also observed in the video footage. Packet 1 (26 vol.%) has a high number of waves breaking on a turbulent flow. As the lahar progresses, surface wave breaking decreases as the flowing mixture assumes a thick, oily, consistency during packet 3 (∼60 vol.%). Remaining wave instabilities have longer wavelengths. The consistency thins throughout packet 4 (∼40%) with an increase in turbulent surface wave breaking through to the tail.

As discussed in “Geophysical recordings of lahars”, the frequency distribution of the seismic signal can provide an indication of the particle concentration and rheological behaviour of the flow, where laminar high concentration flows are commonly characterised by lower frequencies (<<50 Hz) than turbulent hyperconcentrated flows (>50 Hz; Okuda et al. 1979; Suwa et al. 2000; Huang et al. 2004 and 2007; Cole et al. 2009) For this lahar, the largest seismic energies occur between 8–20 Hz, with additional peaks between 28–32 Hz, 40–45 Hz, and the highest at 55–60 Hz (Fig. 3e).

There is little energy in the highest frequencies of packet 1 (Fig. 3e), as is expected for a turbulent, but dilute, flow. In general, there is an increase in the signal amplitude and energy as the particulate volume concentration increases. However, while packet 3 has a larger wetted area (Fig. 3a) and a higher concentration than packet 2 (60 vs. 48 vol.%), it induces lower overall seismic energies (Figs. 3c and e). This may be due to a change to quieter, denser, laminar flow, which should be characterized by less energy in the higher frequencies (Cole et al. 2009). This change to quieter flow may also be due to a change in the grain size distribution (Arattano and Moia 1999). We observe that while packet 3’s seismic response is less energetic; it still has 62% of its energy above 30 Hz. This broad frequency response suggests both low frequency frictional and high frequency collisional processes are occurring in the concentrated phases of these flows.

Estimating particle concentrations

The peak seismic amplitude is commonly correlated to peak flow discharge (e.g. Suwa et al. 2000; Huang et al. 2004). However, Arattano and Moia (1999) suggest that this amplitude is correlated to both flow depth and grain size. Cole et al. (2009) have identified lower peak vibrational energies for laminar flows than turbulent flows. The measured Semeru lahars with cross-sectional areas >2 m2 are positively correlated to the seismic energy (Fig. 4a), showing an increase in seismic energy as the wetted area increases. In addition, the highest concentration flows (≥60 vol.%) are often the flows with the largest wetted areas, and thus the greater seismic energies. However, there is no direct correlation between the concentration and the spectral energy. In addition, the observed power relationship between the wetted area and the spectral energy is approximately the same for dilute, hyperconcentrated, and debris flow type concentrations. Exponents of α = 0.24 ± 0.02 and 0.25 ± 0.02 are found for concentrations of 20–60 vol.% and ≥60 vol.%, respectively (Fig. 4a). Thus, for these flows, this relationship does not provide a useful tool to estimate particle concentrations.
Fig. 4

a Wetted area vs. spectral energy for all events, shown at 10 s intervals after a 20 s smoothing window has been applied. b The averaged sampled concentration in a packet vs. the directionality ratio, which is the ratio of the seismic energy recorded perpendicular to the channel (X-channel), divided by the energy recorded parallel. Error Bars show the range of concentrations in a typical packet (±0.10). c As for (b), but the directionality ratio is divided by the average wetted area

Cole et al. (2009) identified that water-rich flows produce more energetic cross-channel vibrations than flow-parallel, while plug and laminar flows have stronger flow-parallel signals due to low channel-side particle collisions. At Semeru, even the highest concentration lahars exhibit some turbulent behaviour, as indicated by the broad frequency response (see “Field Survey 2008” and “Event 05/03/08”). Thus, the higher concentration packets are expected to have a greater number of turbulent particle collisions. This induces more energy perpendicular to the channel (cross-channel), as illustrated in Fig. 4b. The directionality ratio, which is the ratio between the seismic energy recorded cross-channel and that recorded parallel, increases with increasing concentration. Flows with a larger wetted area will induce additional energy cross-channel, due to the greater surface contact on the channel walls. Figure 4c illustrates a negative correlation between the sampled concentration and the directionality ratio, when divided by the average wetted area for each packet. This negative correlation illustrates the strong relationship between flow size (wetted area) and particle concentration.

These relationships provide a basic tool to estimate concentrations for un-sampled packets. We acknowledge that this needs further verification, as the discrete samples do not capture the full concentration changes throughout the packet. In addition, applying the sampled concentrations from the downstream site to the upstream site assumes that the highest or lowest concentration is associated with the same packet at both sites.

Packet migration

Approximate particle concentrations for each un-sampled lahar ‘packet’ have been calculated from the average of the values estimated by the two relationships shown in Figs. 4b and c. Identification of the packet with the peak concentration is corroborated with video observations (see “Event 05/03/08”). This peak concentration commonly occurs in the packet with the largest wetted area (Fig. 5). Only a few events have their peak velocity associated with the front of the wave. However, historical observations indicate that secondary surges commonly travel slower than the velocity peak at the lahar front. Hürlimann et al. (2003) observed a frontal surge of 4.5 m/s followed by a second surge at 2.7 m/s, while Arattano and Moia (1999) observed frontal velocities of 7.6 m/s, with secondary surges of 7.3 and 6.8 m/s.
Fig. 5

Four example events, showing the wetted area at the upstream ‘lava’ site, packet propagation velocities, and sampled and calculated concentrations corrected to upstream ‘lava’ site arrival times. Vertical grey lines indicate packet arrivals

For the flows recorded at Semeru, the majority of observed events have slower moving frontal packets (1–2 m/s), with the largest, fastest, highest-concentration packet located mid lahar (4–8 m/s, Fig. 5, and “Event 05/03/08”). Lavigne and Suwa (2004) also observed peak velocities occurring some time after the flow front. Superimposed smaller waves are often observed on lahars, travelling at velocities higher than those of the flow itself (see summaries in Arattano and Moia 1999; Massimo 2000). These are often attributed to roll waves (Takahashi 2007; Zanuttigh and Lamberti 2007). However, as discussed in “Field Survey 2008”, the long duration of our recorded packets are not typical of a roll wave phenomenon. We interpret these packets to have originated from multiple spatial and temporal sources, where one meteorological event may have generated a number of dynamically independent lahars which are propagating at different speeds dependent upon their size and particle concentration. These have merged into one single-peak lahar by the time they reach our observation site (“Field Survey 2008”).

The source region is believed to be between 8 to 11.5 km of the observation sites, on the steep flanks of the volcano, where abundant loose material is available for erosion and mobilisation (Fig. 1). Variations in the location of the lahar initiation points within this source region will result in different path lengths, and hence timings, of each discrete lahar packet at our measured reach. Thus, the variety of lahar characteristics seen at our survey site (Fig. 5) may represent different degrees of coalescence of these individual packets within the Lengkong channel, dependent upon the relative source locations and variable velocities between events. These conclusions are validated by the range of lahars we have observed at our study site. These include i) small simple lahars (Fig. 5a), ii) dynamically independent lahars that display only the onset of packet merging (Fig. 5b), and iii) lahars that are composed of multiple discrete events which show increasing degrees of coalescence, becoming more organized as the highly concentrated, faster moving packets catch up with the flow front (Fig. 5c and d).

Eventually, these packets will merge into one large stable lahar. The faster moving packets are also likely to have higher erosive power, and may tend towards debris flow properties as they entrain additional sediment. Marchi et al. (2002) define a ‘debris flow wave’ as a wave of sediment and water with a small hyperconcentrated pre-surge, very shortly followed by a large, steep fronted, debris flow that then dilutes back to a hyperconcentrated flow. The more ‘mature’ lahars recorded at our observation sites, displaying the greatest degree of coalescence (Fig. 5d), may tend towards this ‘debris flow wave’ structure as they continue to propagate downstream, developing a debris flow type rheology with a steep coarse grained flow front. Debris flows such as this have previously been observed at our field site by Lavigne et al. (2003). The variation in the observed debris flow ‘maturity’ may thus depend upon the location of the initiation points, and thus the distance over which the lahar can evolve.

Discussion and conclusions

These initial results indicate that spatial and temporal variations in rainfall induced lahars affect their propagation, evolution and flow behaviour. The fastest portions of the lahar appear to be primarily those with the largest wetted areas, and secondly those with higher concentrations. These high velocity regions are commonly located mid-lahar and not at the flow front, as is commonly seen for mature high concentrated debris flows. The recorded flows may evolve into these debris flows, as they both a) entrain more material, and b) the high velocity, high concentration regions catch up with the flow front. These migrating packets are either consuming the material in front of them, or forcing it to shorten and thicken.

Our results indicate that distinct changes in flow behaviour occur throughout the lahar propagation from source, with multiple peaks and intra-event concentration changes. Single rainfall events may thus generate highly unsteady flows. These can be composed of a series of partly coalesced individual flows, each likely to have been generated from spatially separated tributaries. The original flows continue to behave independently as the lahar moves downstream, until they merge into a single channel to form one multi-peaked lahar. Observations of different degrees of coalescence between these discrete flow packets illustrate that a single mature debris flow may have formed from multiple dynamically independent lahars, each with different origins. These results have applications beyond lahars, to other high energy solid-liquid flows, such as non-volcanic debris flows and flash-floods.

A range of mechanistic models have been developed to simulate lahars. These range from simple empirical based statistical models, such as LAHARZ (Schilling 1998), through to more complex one dimensional dynamic wave theory models of dilute constant concentration sediment-water mixtures (e.g. Macedonio and Pareschi 1992). Two dimensional models, such as FLO-2D (O’Brien 1999) and DELFT3D (see Carrivick et al. 2008) can capture the complex time-varying nature of the sediment-fluid ratio, but are also limited by low sediment concentrations. Models of highly concentrated debris flows commonly adopt a depth-averaged, constant density, pseudo-fluid mixture approach (e.g Iverson and Denlinger 2001). The use of simplifying approximations in these models is important for hazard mitigation, as it allows for large scale simulations over natural terrain with existing computational power. Models that maintain these fast simplifying assumptions, while still being able to incorporate extra complexities such as volume and concentration changes (e.g. Fagents and Baloga 2006), present promising avenues for future development. However, our results indicate that to truly capture the rapid and complicated evolution of these flows, models must not only simulate these types of temporal and spatial concentration changes, but also consider multiple generation sites for each event. Only then can the key changes in flow behaviour be modelled.



We thank Céline Dumaisnil, Yves Bru, the Lengkong villagers, Mahjum and Latif Usman for field assistance, Gert Lube for helpful discussions and Jenny Barclay, Chris Waythomas, and an anonymous journal reviewer for helpful comments to improve our presentation. EED and SJC are supported by the Marsden Fund (MAUX0512) and the NZ FRST (MAUX0401). SEC thanks the Commonwealth Scholarship Scheme and Massey University Graduate Research School. JCT was supported by the French-Indonesian VELI (Volcanisme Explosif Laboratoire Indonésien) research and exchange programme.


  1. Arattano M, Moia F (1999) Monitoring the propagation of a debris flow along a torrent. Hydrol Sci J 44:811–823Google Scholar
  2. Arattano M, Marchi L (2008) Systems and sensors for debris-flow monitoring and warning. Sensors 8:2436–2452CrossRefGoogle Scholar
  3. Beverage J, Culbertson J (1964) Hyperconcentrations of suspended sediment. Am Soc Civ Eng 90:117–126Google Scholar
  4. Carrivick JL, Manville V, Cronin SJ (2008) A fluid dynamics approach to modelling the 18th March 2007 lahar at Mt. Ruapehu, New Zealand. Bull Volcanol 71:153–169. doi:10.1007/s00445-008-0213-2 CrossRefGoogle Scholar
  5. Cole S, Cronin S, Sherburn S, Manville V (2009) Seismic signals of snow- slurry lahars in motion: 25 Sept. 2007, Mt. Ruapehu, N.Z. Geophys Res Lett 36:L09405. doi:10.1029/2009GL038030 CrossRefGoogle Scholar
  6. Coussot P, Meunier M (1996) Recognition, classification and mechanical description of debris flows. Earth Sci Rev 40:209–227CrossRefGoogle Scholar
  7. Cronin SJ, Neall VE, Lecointre JA, Palmer AS (1997) Changes in Whangaehu River lahar characteristics during the 1995 eruption sequence, Ruapehu volcano, N.Z. J Volcanol Geotherm Res 76:47–61CrossRefGoogle Scholar
  8. Cronin SJ, Neall VE, Lecointre JA, Palmer AS (1999) Dynamic interactions between lahars and stream flow: a case study from Ruapehu volcano, N.Z. Geol Soc Am Bull 111:28–38CrossRefGoogle Scholar
  9. Doyle EE, Cronin SJ, Cole SE, Thouret J-C (2009) The challenges of incorporating temporal and spatial changes into numerical models of lahars. In: Proceedings of the 18th World International Congress on Modelling and Simulation, Modelling and Simulation Society of Australia and New Zealand, Cairns, 13–17 July 2009Google Scholar
  10. Dumaisnil C, Thouret J-C, Chambon G, Doyle EE, Cronin SJ (2010) Distinctive hydraulic characteristics and a frictional model apply to lahar flows at Semeru volcano (Indonesia). Earth Surf Proc Landf. doi:10.1002/esp.2003
  11. Fagents SA, Baloga SM (2006) Toward a model for the bulking and debulking of lahars. J Geophys Res 111:B10201. doi:10.1029/2005JB003986 CrossRefGoogle Scholar
  12. Huang CJ, Shieh CL, Yin HY (2004) Laboratory study of the underground sound generated by debris flows. J Geophys Res 109:F01008. doi:10.1029/2003JF000048 CrossRefGoogle Scholar
  13. Huang C-J, Yin H-Y, Chen C-Y, Yeh C-H, Wang C-L (2007) Ground vibrations produced by rock motions and debris flows. J Geophys Res 112:F02014. doi:10.1029/2005JF000437 CrossRefGoogle Scholar
  14. Hürlimann M, Rickenmann D, Graf C (2003) Field and monitoring data of debris-flow events in the Swiss Alps. Can Geotech J 40:161–175CrossRefGoogle Scholar
  15. Iverson RM (1997) The physics of debris flows. Rev Geophys 35:245–296CrossRefGoogle Scholar
  16. Iverson RM, Denlinger RP (2001) Flow of variably fluidized granular masses across three-dimensional terrain. 1. Coulomb mixture theory. J Geophys Res 106:537–552CrossRefGoogle Scholar
  17. Iverson RM, Reid ME, LaHusen RG (1997) Debris-flow mobilization from landslides. Ann Rev Earth Planet Sci 25:85–138CrossRefGoogle Scholar
  18. Lavigne S, Suwa H (2004) Contrasts between debris flows, hyperconcentrated flows and stream flows at a channel of Mt. Semeru, East Java, Indonesia. Geomorphology 61:41–58CrossRefGoogle Scholar
  19. Lavigne F, Thouret J-C, Voight B, Young K, La Husen R, Marso J, Suwa H, Sumaryono A, Sayudi DS, Dejean M (2000) Instrumental lahar monitoring at Merapi volcano. J Volcanol Geotherm Res 100:457–478CrossRefGoogle Scholar
  20. Lavigne F, Tirel A, Le Froch D, Veyrat-Charvillon S (2003) A real-time assessment of lahar dynamics and sediment load based on video-camera recording at Semeru volcano, Indonesia. In: Rickenman D, Chen C (eds) Debris-flow hazards mitigation: mechanics, prediction and assessment, vol 2. Millpress, Rotterdam, pp 871–882Google Scholar
  21. Macedonio G, Pareschi M (1992) Numerical simulation of some lahars from Mount St. Helens. J Volcanol Geotherm Res 54:65–80CrossRefGoogle Scholar
  22. Major JJ (1997) Depositional processes in large-scale debris-flow experiments. J Geol 105:345–366CrossRefGoogle Scholar
  23. Manville V, Cronin SJ (2007) Breakout lahar from New Zealand’s Crater Lake. EOS Trans Am Geophys Union 88(43):441–456CrossRefGoogle Scholar
  24. Marchi L, Arattano M, Deganutti AM (2002) Ten years of debris-flow monitoring in the Moscardo Torrent (Italian Alps). Geomorphology 46:1–17CrossRefGoogle Scholar
  25. Marcial S, Melosantos AA, Hadley KC, LaHusen RG, Marso JN (1996) Instrumental lahar monitoring at Mt. Pinatubo. In: Newhall CG, Punongbayan RS (eds) Fire and mud: eruptions and lahars of Mt. Pinatubo, Philippines. Seattle, University of Washington Press, pp 1015–1022Google Scholar
  26. Massimo A (2000) On debris flow front evolution along a torrent. Phys Chem Earth (B) 25:733–740Google Scholar
  27. O’Brien JS (1999) FLO-2D users manual, version 99.2, pp 157Google Scholar
  28. Okuda S, Suwa K, Okunishi K, Yokoyama K, Ogawa K, Hamana S (1979) Synthetic observation of debris flows (part 5). Annuals of Disaster Prevention Research Institute, Kyoto University, 22:175–204. Japanese with English Abstract.Google Scholar
  29. Pierson TC (2005) Hyperconcentrated flow transition process between water flow and debris flow. In: Jakob M, Hungr O (eds) Debris flow hazards and related phenomena. Springer-Verlag, New York, pp 159–196CrossRefGoogle Scholar
  30. Rickenmann D, McArdel BW (2007) Continuous measurement of sediment transport in the Erlenbach stream using piezoelectric bedload impact sensors. Earth Surf Proc Landf 32:1362–1378. doi:10.1002/esp.1478 CrossRefGoogle Scholar
  31. Schilling SP (1998) LAHARZ, GIS programs for automated mapping of lahar-inundation hazard zones. US Geol Surv Open-File Rep 98–638:1–79Google Scholar
  32. Suwa H, Yamakoshi T, Sato K (2000) Relationship between debris-flow discharge and ground vibration. In: Wieczorek G, Naeser N (eds) Debris-flow hazards mitigation: mechanics, prediction and assessment. Millpress, Rotterdam, pp 311–318Google Scholar
  33. Takahashi T (ed) (2007) Debris flow: mechanics, prediction and countermeasures. Taylor & Francis, LondonGoogle Scholar
  34. Witham C (2005) Volcanic disasters and incidents: A new database. J Volcanol Geotherm Res. 148:191–233. doi:10.1016/j.jvolgeores.2005.04.017 CrossRefGoogle Scholar
  35. Zanuttigh B, Lamberti A (2007) Instability and surge development in debris flows. Rev Geophys 45:RG3006. doi:10.1029/2005RG000175 CrossRefGoogle Scholar
  36. Zobin VM, Plascencia I, Reyes G, Navarro C (2009) The characteristics of seismic signals produced by lahars and pyroclastic flows: Volcán de Colima, México. J Volcanol Geotherm Res 179:157–167CrossRefGoogle Scholar

Copyright information

© Springer-Verlag 2010

Authors and Affiliations

  • E. E. Doyle
    • 1
  • S. J. Cronin
    • 1
  • S. E. Cole
    • 1
  • J.-C. Thouret
    • 2
  1. 1.Institute of Natural ResourcesMassey UniversityPalmerston NorthNew Zealand
  2. 2.Laboratoire Magmas et Volcans, UMR 6524 CNRSUniversité Blaise PascalClermont-FerrandFrance

Personalised recommendations