Recent Landslides Landslides (2010) 7:303–315 DOI 10.1007/s10346-009-0180-5 Received: 14 June 2009 Accepted: 24 September 2009 Published online: 31 October 2009 © Springer-Verlag 2009

Christopher I. Massey . Vernon Manville . Graham H. Hancox . Harry J. Keys . Colin Lawrence . Mauri McSaveney

Out-burst flood (lahar) triggered by retrogressive landsliding, 18 March 2007 at Mt Ruapehu, New Zealand—a successful early warning

Abstract The summit crater of Mt Ruapehu volcano normally hosts a 15.4-ha warm lake, whose water has been repeatedly wholly or partly ejected by explosive and extrusive eruptions. Some of the larger eruptions have modified the lake outlet by burying it under unconsolidated tephra (volcanic ash and blocks), creating a dambreak flood hazard independently of the occurrence of an eruption. Eruptions in 1995 and 1996 followed this sequence; a break-out flood was anticipated and a warning system was installed to mitigate the risk from this event and subsequent lahars in the same catchment. The 11-year filling time allowed much planning and rehearsal. The warning system involved manual inspections of dam integrity, and seepage and lake-level monitoring to constrain the likely failure window, and telemetered instruments including a tripwire and geophones to detect breaching of the dam and propagation of the outbreak flood. The dam-collapse sequence, captured by a time-lapse camera, involved a series of retrogressing landslides initiated and accelerated by seepage forces and toe scour when the lake was 1.1 m below overtopping. The barrier failed in two phases on 18th March, 2007, beginning at 09:55 (NZST), with rapid retreat of one of the erosion scarps on the downstream slope of the eastern barrier, initiated by internal erosion. Headward retrogression of the scarp into the barrier formed an initial breach in the dam, after which increasing outflow led to erosion and undercutting of the wider downstream toe of the western barrier. A final, larger dam breach occurred between 11:21 and 11:22 as slope instability caused retrogressive failure of the remaining barrier. Five-hundred meters downstream of the dam, a large landslide was reactivated by toe scour during the flood, contributing about a million cubic meters of solid material to the volumetric bulking of the outflow, which reached the coast, 215 km away, 17 h later. The success of the planning and warning system allowed the whole event to occur with little damage to infrastructure and without causing injury. Keywords Natural-dam collapse . Landslide retrogression . Break-out flood . Lahar . Tephra barrier . Landslide warning . Crater lake Introduction The issuing of early warnings of the imminent hazard of landslides triggered by heavy rain and increased phreatic pressures in unstable slopes is seen in many countries around the world as an effective means of reducing loss of life. In this example from a recent event in New Zealand, we discuss an effective early warning and the causes of retrogressive landslide failures in a natural volcanic dam, whose failure triggered a much larger secondary geohazard than the initial slope collapses. This example is the Eastern Ruapehu Lahar Warning System (ERLAWS), and the failure of the tephra dam by retrogressive landsliding that led to the Ruapehu Crater Lake outburst flood (lahar) on 18 March 2007.

Mt Ruapehu (2,797 m1) is an active andesitic stratovolcano in the central North Island, New Zealand (Fig. 1). The active crater hosts a 15.4-ha, warm, acidic Crater Lake above the perennial snowline (Christenson and Wood 1993): volcanogenic floods (or lahars) have thus accompanied most significant historic eruptions due to the explosive ejection of lake water onto the summit slopes of the mountain or displacement over the normal lake outlet (Cronin et al. 1997b). In addition, lahars have been triggered by heavy rainfall remobilizing tephra deposits in the aftermath of magmatic eruptions (Hodgson and Manville 1999), and by the breaching of a temporary barrier of volcanic and other debris emplaced over the outlet area (Manville et al. 2007a, b). One such break-out, in 1953, 8 years after an eruption sequence that emptied the lake, caused a lahar that killed 151 people in New Zealand’s worst volcanic disaster (Manville 2004; O’Shea 1954). In 1995–96, the largest eruption sequence in 50 years progressively emptied the lake (Johnston et al. 2000; Manville et al. 2007a, b), generated a series of eruption- and rain-triggered lahars, and constructed a 7–10-m thick barrier of loose tephra at the lowest point in the crater rim (Figs. 2 and 4). The former outlet area had comprised a lava rock ledge that had controlled lake overflow since 1953 (Hancox et al. 1997). In the following 11 years, the about 9 million-m3 lake refilled at an irregular rate from a combination of precipitation and juvenile inputs, mediated by the seasonal alpine climate and geothermal heating cycles, raising the possibility of a repeat of the 1953 disaster. However, society was forewarned this time, resulting in the planning and implementation of a series of mitigation measures (Keys and Green 2008). The implemented measures included manual monitoring of both the lake level to constrain the likely time of failure (Gillon et al. 2006), and the integrity of the dam in order to detect early signs of weakening. This was backed up instrumentally by what became known as the ERLAWS (discussed in detail in Keys and Green 2008), which included sensors designed to detect the initial failure of the dam (locations shown in Fig. 3). Independently, a scientific research plan was developed (Manville and Cronin 2007), which included additional sensors at the crater rim including a bubblein lake-level sensor and a time-lapse digital still camera. In part, an objective of this paper is to record for the landslide specialist, a successful use and outcome of the warning system. Although in New Zealand Law, a lahar that has the consistency of a debris flow is treated for legal purposes as a landslide (falling debris), we do not especially deal with the lahar, which has been documented elsewhere and is the subject of much ongoing research (Manville and Cronin 2007; Carrivick et al. 2009; and other manuscripts in preparation). Instead, this paper focuses on the sequence of minor landslides that led to failure of the tephra barrier, the partial release of Crater Lake, and a number of 1 All

heights are given in metres above mean sea level datum Landslides 7 • (2010)

303

Recent Landslides Fig. 1 Location of Mt Ruapehu, New Zealand

35˚

N

Auckland

Mt Ruapehu

SS 40˚

Wellington

Christchurch

45˚

170˚ 170˚E E

landslides triggered along the outflow channel by toe scour, including a 106 m3 reactivation of the Whangaehu landslide.

The planning phase The central and local government and infrastructural and other agency preparations for the forecast break-out lahar focused on what was needed to mitigate the off-site lahar hazard (Keys and Green 2008). Planning took place within New Zealand’s unique cultural, administrative, and legislative environment, with full and open access to the world’s scientific and engineering expertise. The decisions made and implemented were well suited to these environments and the situation at hand, and thus may not suit or be applicable to other social settings and contexts. Implementation was completed well before the expected event occurred although the slower than anticipated lake filling rate allowed additional work and more thorough training of response agencies. Planning recognised that the expected lahar in the Whangaehu River would essentially be a recurrent, although infrequent event, rather than a unique one. It also recognized the other ways lahars 304

Landslides 7 • (2010)

175˚E 175˚

occur in this valley, including very large lahar/landslide events which have occurred as recently as 400 years ago (Hodgson et al. 2007) and potentially from a large failure of the aging crater rim. The New Zealand hazard-management environment favours resilience to hazard occurrences rather than their prevention, hence it was resolved that if mitigation were to be contemplated, it ought to present a long-term solution (and not merely an immediate, “one-off” solution). This pointed to measures akin to conventional flood mitigation measures, rather than treating the particular developing event at source through engineering intervention. Effectively, the forecast break-out was treated as a flood in the areas subject to a flooding hazard, rather than as a problem to solve at source. This scientifically based rationale had a number of administrative advantages in the New Zealand social environment: (a) It did not interfere in any significant way with the natural values and purpose of the Tongariro National Park where outstanding natural volcanic values are recognised in its World Heritage status.

(b) It did not interfere with the summit area of the volcano, which was viewed by some as sacred, and by others as a dangerously dynamic and challenging operating environment in which to manage risk, complicated by the high altitude and often severe weather. (c) It would have minimal impact on the natural environment in which the occurrence of lahars, which often occur without warning, was an expected part. In addition, the at-risk river, the Whangaehu, is already recognized as New Zealand’s most polluted waterway due to natural contamination by geothermal sulphur and heavy-metal compounds, and flows through a remote area with little infrastructure or population to protect. (d) It put the major mitigation–implementation costs directly on the components of New Zealand society placed at risk, and the work and costs were feasible for the components involved. (e) If it worked successfully this time for a lahar which was predicted to be larger that the 1953 event (Hancox et al. 2001), it might work again several times in the future with little additional cost. (f) It also covered the possibility of an eruption-triggered lahar, the residual risk of which was not affected as long as there was a significant volume of water in the lake, even below overflow. It had no obvious disadvantages other than it left the event to occur in its own time, and not predictably on demand. This was not viewed as a compelling reason for going against any of the other reasons, and the uncertainty in precise timing could be well covered by a reliable warning system. Previous attempts at engineered mitigation of volcanogenic break-out floods have met with mixed success (Bornas and Team 2002; Bornas et al. 2003) and did not form a strong precedent for intervention in this case.

Fig. 2 Views of the tephra barrier at Crater Lake before and after failure

The warning system A detailed account of the ERLAWS and its performance is given in Keys and Green (2008). The intention was to provide a variety of independent signal pathways to provide back-up for failure of any

Fig. 3 View of the barrier before failure showing locations of the monitoring instrumentation

AFM, radio, power supply Water pressure transducers

Geophone 1.3 Geophone 1.2

T

R

IP

W

IR

E

Geophone 1.1

Landslides 7 • (2010)

305

Recent Landslides one pathway, but still provide independent corroboration of the presence of a lahar in the Whangaehu channel, in near real-time, to those with a need to respond. The main physical components comprised three independent monitoring sites, one at the crater rim to monitor the integrity of the tephra dam (Site 1), and two further sites at 2.2 and 4.6 km downstream to provide serial detection of the passage of an outbreak flood resulting from dam failure, or an eruption-triggered lahar. Instrumentation was designed by the United States Geological Survey (instrument specifications are given in http://pubs.usgs.gov/sir/2008/5114/ sir2008-5114.pdf). Data were relayed via radio-telemetry to a base station with capability to automatically transmit an alarm to key agencies with rostered staff providing 24-h, 7-days-a-week response capability via pagers, telephones and the internet (see “Time-varying likelihood of failure (conditional probability)” section below). Site 1 comprised three geophones buried in the dam to detect vibrations associated with the onset of failure and/or outflow, and a tripwire to detect physical breaching of the dam crest. When transmitted signals exceeded a pre-set threshold voltage (500 mV for the geophones, predetermined by appropriately setting the instrument amplifier gain) alarms were sent in near real-time to Police and other agencies. When two or more acoustic flow monitors triggered within a prescribed time limit, a specific alarm was sent with high priority indicating the probable occurrence of a lahar. In such situations, flashing lights were activated on a major highway where it crossed potential overspill channels, while lights and automatic barrier arms were used to close a road bridge over the Whangaehu River 39 km from Crater Lake. The adjacent rail bridge was also closed to trains until such time as it had been inspected for possible damage. In addition to the ERLAWS system, equipment designed to detect potential lahar threats to their assets were operated by two infrastructure companies, hydrological monitoring sites were operated by a local regional council, and numerous temporary sensor arrays were installed by a diverse group of researchers from GNS Science in collaboration with New Zealand and overseas universities (Manville and Cronin 2007). Although all this attention was focused on the Whangaehu River, the natural outlet to Crater Lake, lahars have been triggered in multiple other catchments by explosive eruptions (Nairn et al. 1979) or rainremobilisation of ash (Hodgson and Manville 1999; Manville et al. 2000). Very large prehistoric lahars in the Whangaehu have also spilled northwards into the Tongariro River catchment (Cronin et al. 1997a): some of these likely resulted from failures of the hardrock rim of the crater basin (Hodgson et al. 2007; Lecointre et al. 2004). An independent seismic network operated by GNS Science (GeoNet), continuously monitors for volcanic events, and has the capability to detect a large rim collapse, and possible precursory earthquakes. Many of these systems were optimized for the detection of large, sediment-laden lahars. Small-volume ice-slurry lahars triggered by a phreatic eruption through Crater Lake on 25 September 2007 were detected by these systems (Kilgour et al. 2009; Lube et al. 2009) but not at sufficient resolution or amplitude to trigger an ERLAWS warning. An additional lahar damaged some skifield infrastructure elsewhere on the mountain, and narrowly missed some skifield staff. These events fell into the category of residual risk, which planning had considered. 306

Landslides 7 • (2010)

Time-varying likelihood of failure (conditional probability) Planning for the eventual Crater Lake break-out recognized that the interval when the tephra dam could breach was not a completely random variable, but became more likely the higher the lake rose (this is technically called conditional probability). Early investigations showed that the lava sheet forming the previous lake outlet survived the 1995–96 eruptions although it had been eroded by up to 0.7 m: therefore dam failure was a negligible danger until the lake had refilled above the base of the tephra dam (2,539.3 m), at which point its likelihood would begin to increase dramatically. Until such time, the only hazard was the residual risk of a volcanic eruption. As a consequence of this foresight, an integral part of the ERLAWS system was the establishment of a series of warning (alert) levels (Table 1) based on the water level of the Crater Lake relative to the old, and now buried outlet. These warning levels were used to guide resource allocation (especially speed of response) and budgeting within the various emergency agencies as the likely time to failure approached. Initially, in 2002, these alert levels were based on the then-existing understanding of the anticipated lahar hazard in relation to critical lake levels and lake fullness based on the lake level–volume relationship. The suggested response times could only be indicative at the beginning. To account for the significant uncertainty in some lake levels, some of the values used were allowed to evolve as acquisition of new data provided more reliable information and prediction. In 2003, as a result of further investigation of a hard-rock rim buried under the tephra dam, warning levels were based on the post-1995 level of the hard-rock rim of the Crater Lake at the outlet. At this stage, lake-fullness values were expressed as a proportion (%) of the volume contained by the post 1995 hard-rock rim (100% indicated that the lake had reached the base of the tephra dam). As the lake exceeded this rim level, however, it became obvious that proportions larger than 100% meant little to many of the agencies or to the public. As a consequence, in 2005, the reference for threshold values was changed to absolute lake level (as used in Table 1 and various Figures), based on heights of the lake-water surface above or below the level of the rock rim. So-called “critical levels” at which alert levels changed were based on historical records of previous lahars, the damage they caused, their magnitude, and their relative travel times. Also considered was the estimated relative probability of such a lahar occurring at each level. These levels were still estimates however and were intended to be guides in recognition that no certainty could be given. Similarly, the potential thresholds for onset of risk to structures were estimated based on historical records and some likely damage thresholds previously calculated, but they were never linked to specific engineering assessments of damage thresholds. This lake-level-based warning-level system was used in all the different agency response plans for staff readiness (i.e. speed of response), proximity and resource allocation. In the Department of Conservation, for example, this included the frequency of monitoring the lake and dam, ERLAWS readiness and testing frequency, numbers of emergency staff allocated to stand-by, public-awareness work, and sign deployment (and consequent necessary increases in expenditure). The system was fully documented in an appendix to the Department of Conservation Eastern Ruapehu Lahar Emergency Response Plan and appeared in outline form in all other agencies’ plans.

2,536 m

2,536.5

2,536.9

Level 3b

Level 4

Level 5

Response within 5 min

2,535 m Level 3a

Equivalent to a large, fast lahar. Conditional probability is 90% Lake at top of the tephra dam. Conditional probability is 100%

2,533 m Level 2

Response within 10 min

0.7 to 1.9 months to fill from Level 2 to 3, or 7.8 months to drop to Level 2 from Level 3, depending on infill rates. Variation due to the possibility of filling spanning fast and slow filling rates, and seepage. Slow fill rates will probably result in net drops in lake level above about 2,532 m 0.4 to 0.6 months to fill from level 3 to 3b, or 3.2 months to drop to level 3 from 3b 0.2 to 0.3 months to fill from level 3b to 4, or 1.1 months to drop to level 3b from 4 0.2 to 0.3 months to fill from level 4 to 5, or 0.7 months to drop to level 4 from level 5 Response within 20 min (for example, this required one local police sergeant to always be within 20 min of base from this time on)

2,529.5 m (Lake 100% full) Level 1b

Below 2,527 m 2,526.5 m

Equivalent to a large moderately fast lahar. Conditional probability of dam failure at this level is 5–10% Conditional probability is 50–60%

1 to 6 months to fill from Alert Level 1b to 2 Planning completed. Full response capability available and ready

1 to 6 months to fill from Alert Level 1 to 1b Planning, preparation, and training Planning largely completed. Response capability available. Response within 30 min

Anticipated time for lake to rise to next level (in summer, based on fill rates 2000–2005) Actions (mostly agency response time)

Simplified explanation (conditional probability of dam failure based on Gillon et al. 2006) Base level of readiness as per normal civil defence planning Critical trigger point, 3 m below the new rock overflow level. (Waves caused by small eruptions or landslides could overtop tephra barrier but resulting lahar would be small) Lake reaches to the buried rock rim outlet level at the base of the tephra dam. Probability of dam failure at this level is still very low Sudden collapse could produce a lahar equivalent to the 1975 event (largest historic eruption lahar) which passed under downstream road and rail bridges without significant damage. Conditional probability of dam failure at this level is 1–2% Lake level (msl)

Level of readiness (warning level) Normal Level 1

Table 1 The ERLAWS warning (alert) levels—basis for system, general response time, actions, and expected time between levels

Investigations of potential dam-failure modes indicated that once the barrier began to breach, possibly by overtopping, outflow would continue at an increasing rate due to erosion and scour, leading to a catastrophic release of water rather than gradual seepage until the excess head in the lake was drawn down. The most likely time that ERLAWS would activate was during the interval when the lake level was above the former lake outlet: a state reached briefly for the first time in January 2005. To monitor this, lake level was measured during site visits and automatically recorded and telemetered, although this did not form part of the break-out detection algorithm. High lake temperatures, very acid water, waves in the lake induced by ice calving, small eruptions and bank collapse all created problems for the lake-level monitoring equipment and made continuous automatic monitoring very unreliable. The chemical composition of water in the channel below the dam was monitored for chloride concentrations and sulphate/ chloride ratios to detect Crater Lake water contributions to the otherwise glacier and snow-melt fed stream. Leakage from the lake was first detected in May 2006 at a lake level of 2,530.9 m. This monitoring continued until the end of 2006 when visible seepage could be observed through holes in the 2006 Winter–Spring snow cover. Then, as access to the seeps on the outside face of the dam became possible, the actual flow was measured using a V-notch weir from January until 4 March 2007. Geophysical monitoring also detected saturation of the tephra dam by electrically conductive mineralized Crater Lake water (Turner et al. 2007). Several methods of automated seepage monitoring were considered but were not assessed to be cost-effective or reliable enough, and plans to install borehole piezometers ran out of time. The frequency of direct observations of the lake level and the dam were increased as the various lake-level thresholds were exceeded, and as evidence of minor slope failures at the seepage sites became apparent. Simultaneously, erosion of the lake-ward face of the tephra dam occurred due to a combination of wave action and hydraulic sapping during level rise. These stages were recognized in the response plan (Table 1), with the various response agencies raising their alert levels as appropriate. The tephra barrier geological model A geological model of the tephra barrier was first developed by Hancox et al. (1997 and 2001), based mainly on surface geological mapping. More detailed investigations, including dynamic conepenetrometer (DCP) testing, test pitting, sampling, particle-size analysis and ground-penetrating radar (GPR) surveys were carried out jointly by GNS Science and the Department of Conservation (DOC), on behalf of the Crater Lake Scientific and Technical Advisory Panel (STAP) set up by DOC (Keys 2003), and were used to modify the model, (Keys 2003 and Gillon et al. 2006). The model was further refined in 2007 based on the results of topographic surveying using terrestrial laser scanning (TLS) surveys and detailed field mapping pre- and post-failure. A location map and cross sections through the tephra barrier are shown in Figs. 4 and 5, respectively. Detailed logging by Houghton (2007, personal communication) showed the existence of 17 different material layers in the barrier, which could be grouped into two main types, based on geotechnical properties in Gillon et al. (2006). These materials were: (1) Fine tephra (tephra)— fine to coarse SAND, with 2% to 10% fine gravel and 5% to 13% silt; and (2) Coarse tephra (Lapilli)—fine to coarse GRAVEL) with 1% to Landslides 7 • (2010)

307

Recent Landslides B

CRATER

LAKE

A

N

IT7

W6

E1

W4

E1 HOT

W1

seepage location C0

E3 E4

Geophone 1 ICE 10 m

0

A’

Dam camera

B’

KEY 1-m contours derived from laser scanning carried out 4/3/2007

E4

Head of erosion scarps recorded 4/3/2007

Mt Ruapehu Crater Lake

Seepage location (evidence of sub-surface water flow) Section lines (A-A’, B-B’) Lake level at failure (2535.8) Survey mark IT7

Area of figure above

Extent of Ice recorded

Fig. 4 Map of the tephra barrier, on the southern edge of the active vent of Mt Ruapehu

7% silt. Both materials were cohesionless, and based on the DCP testing, the lapilli has a higher relative density (dense to very dense) than the other tephra (medium dense), (NZGS 2005). The lapilli formed two distinct laterally continuous layers within the barrier. The lower lapilli layer varied in thickness, typically 0.5–1.0 m, and outcropped at the toe of the downstream slope of the barrier. The upper lapilli layer was thicker, typically 1.5–2.0 m, and while it extended into the upstream slope of the barrier, it did not crop out on the downstream slope. The tephra and lapilli layers overlay lava from pre-1945 eruptions, with the base of the tephra from 1995–96 eruptions 308

Landslides 7 • (2010)

determined originally from DCP testing (Keys 2003) and updated from TLS surveys carried out after barrier failure. There was agreement between the level estimated prior to the barrier failure (2,529.3±0.5 m AMSL, Keys 2003), and that derived from TLS surveys carried out post barrier failure (2,529.3±0.2 m AMSL). Observations made during lake filling The level of Crater Lake was routinely monitored by DOC, supported by GNS Science. The first lake-level survey was on 21 September 1996 and the last on 19 March 2007, the day after the breach, although less

A

A’

North

South Pre- failure ground surface (4 Mar 07)

Lake level at failure 2535.8

Crown of erosion scarp E1 8 Feb 07 Crown of erosion scarp E1 18 Jan 07 Seepage point E1 hot (2531.7) Ice 18 Jan 07

Lake level post failure 2529.4

Ice 8 Feb

Post failure ground surface (2529.3)

Ice 4 Mar Ice 23 Mar

B

B’

North

South

18 MAR 07 Lake level at failure 2535.8

Pre- failure ground surface (4 MAR 07)

Head of erosion scarp CO CO Seepage point (2529.9) Ice 18 Jan 07

Lake level post failure 2529.4

Ice 8 Feb 07

Post failure ground surface

10 m

(V=H)

Ice 4 Mar 07 Ice 23 Mar 07

KEY

Tephra (medium dense, slightly silty SAND) Lapilli (dense to very dense, slightly sandy GRAVEL Lava (very strong, ANDESITE) (Material description NZGS, 2005)

Water / Ice Material boundary (derived from scala penetrometer testing and field mapping) Ground surface 8 Feb 07 caused by retrogression of erosion scarp E1

Pre-failure ground surface (derived from laser scan survey (4 Mar 07) Post-failure ground surface (derived from laser scan survey (23 Mar 07) Water level (lake level and groundwater)

Fig. 5 Cross-sections through the tephra barrier (locations shown on Fig. 4)

frequent monitoring has continued. Measurements were made one to several times each month. Surveys used an Abney level referenced to stable survey marks located off the tephra barrier, which were checked annually using a total survey station. On 27 January 2006, the lake again reached the elevation of the lava lip that formed the overflow channel prior to the 1995– 96 eruptions, and the base of the tephra (and lapilli) forming the new barrier, triggering the raising of the ERLAWS warning level to 1b (Table 1). The lake level formed an important part of ERLAWS, as the warning levels were based on particular lake levels based on a probabilistic assessment of various failure scenarios (Gillon et al. 2006). Figure 6 shows lake levels and associated alert levels between 1 October 2006 and 19 March 2007. As the lake continued to rise, additional GNS Science monitoring equipment was installed on the margins of the lake, including a bubbler system logging lake level to ±3 mm precision at 10-s intervals, and an automated digital still camera overlooking the downstream slope of the tephra barrier, storing images to an on-site flashcard at 1-min intervals during daylight hours. When the lake level was at 2,534.2 m, areas of seepage and localised erosion on the downstream slope of the tephra barrier were identified between 24 December 2006 and 5 January 2007

(erosion scarps C0, E1, E3, E4, W1, W4 and W6, Fig. 4, following melt-back of seasonal snow from the dam toe). To quantify the rate and extent of the erosion, a series of TLS surveys of the barrier were carried out along with other field measurements, commencing on 18 January 2007, with additional surveys on 8 February, 4 March, and finally on 23 March 2007, 5 days after failure. Difference models extracted from the TLS survey data were used to examine the evolution of the erosion scarps, which indicated northerly retrogression into the barrier, upslope towards the dam crest and lake. The maximum retrogression was 4.7 m towards the NNE, recorded between 18 January and 8 February 2007 for erosion scarp E1 (Fig. 7), with only minor retrogression (0.5 m) of the same scarp recorded between the 8 February and 4 March surveys. The period of maximum retrogression coincided with a period of lake rise from 2,534.9 to 2,535.5 m, whereas the minor retrogression was recorded during a period of relative lake stability (2,535.4–2,535.5 m). Observations made at the same time as the TLS surveys indicated that the erosion appeared to have initiated through confined seepage within the lower layer of coarse tephra (lapilli), outcropping on the downstream slope (Fig. 5). At the time of the 4 January inspection, seepage through this flow path was eroding and transporting material from the overlying fine tephra (Fig. 5), Landslides 7 • (2010)

309

Recent Landslides 2540 Failure of barrier 18 MAR 07 @11:22 (NZST)

Crater Lake Level m AMSL

2538

2536

24 DEC 06 18 JAN 07 Seepage through 1st laser barrier first identified scan survey

8 FEB 07 2nd laser scan survey

4 MAR 07 3rd laser scan survey

23 MAR 07 post failure laser scan survey

ERLAWS warning level 3b ERLAWS warning level 3a

2534 ERLAWS warning level 2

2532

2530

Lake level

Lava Lip (2529.3 m AMSL)

ERLAWS warning level 1b

20/09/06

10/10/06

30/10/06

19/11/06

09/12/06

29/12/06

18/01/07

07/02/07

27/02/07 19/03/07

Fig. 6 Crater Lake levels between October 2006 and barrier failure on 18 March 2007 (from Abney level surveys)

which was in turn leading to localised slumping and development of the erosion scarps on the downstream slope. Active erosion was also observed on 18 January, however, on 8 February and 4 March 2007 water seeping from the lapilli layer was relatively clear and active erosion was not apparent. Barrier-failure sequence On Sunday 18 March 2007, the tephra barrier breached following a prolonged period of wet weather. The lake rose from 2,535.4 to 2,535.7 between 13 and 15 March following a period of heavy rain (cumulative rainfall about 780 mm over 2 days, measured 4.5 km

CRATER B

east of Crater Lake). On the morning of 18 March 2007, a rainstorm (cumulative rainfall about 256 mm over 10 h) caused the lake to rise from 2,535.7 to 2,535.8 m (Fig. 8). Despite poor weather, the fixed camera captured a time-lapse sequence of the failure (the lowresolution images are viewable at www.gns.cri.nz/news/release/ 20070327_lahar_words.html), while lake drawdown was obtained from both the bubbler and ERLAWS lake-level sensors. Meanwhile ERLAWS geophones at the tephra barrier detected vibrations caused by the sequential collapse of the tephra dam and the escaping water (Keys and Green 2008). These data were used to infer the barrierfailure sequence (Table 2 and Figs. 9 and 10).

KEY

LAKE

N

A

1-m contours derived from TLS survey 8 Feb 2007

E4 Lake level at failure (2535.8)

IT7

W6

Lake level at failure (2535.8)

W1

Survey mark IT7

C0

Extent of Ice recorded 8 Feb 2007

E3 E4

ICE A’

Seepage location (evidence of subsurface water flow) Section lines (A-A’, B-B’)

E1

W4

Head of erosion scarps recorded 4/3/2007

Geophone 1 B’

Dam camera

Surface change Jan 18 to Feb 8 2007 Low : 0

0

10 m

High : 9.5m

Fig. 7 Map of surface change derived from the difference model generated from terrestrial laser scan surveys made on 18 January 2007 and 08 February 2007

310

Landslides 7 • (2010)

18/03/07 Lake level 2535.8

15/03/07 Lake level 2535.7 Lake level 2535.4

2536

No data

0.8

0.7

2535

Lake level (m AMSL)

0.9

0.6 2534 Bubble data

0.5

ERLAWS

2533

Rainfall 0.4

2532 0.3

Rainfall (mm/5-minute interval)

2537

2531 0.2 2530

0.1

2529 11/03/07

12/03/07

13/03/07

14/03/07

15/03/07

16/03/07

17/03/07

18/03/07

19/03/07

0 20/03/07

Fig. 8 Crater Lake level (from bubbler sensor and the ERLAWS water-level sensor) and rainfall (from Tukino rain gauge, 4.5 km East of Crater Lake)

Barrier-failure mechanisms Relationships between the time-lapse photographs, intensity of the geophone signals, and lake drawdown indicate that the barrier failed sequentially, and involved two main failure mechanisms. The rapid failure sequence appears to have been initiated by the increased seepage forces caused by the rising lake, combined with direct infiltration of rain on the dam from the rainstorm on the morning of

18 March 2007. The failure sequence began at 09:55 (NZST), by rapid retreat of the eastern flank of erosion scarp E1, at the approximate location of seepage point E1-hot (Fig. 4). Water escaping from E1-hot had been consistently warmer than at any other seepage point, and is inferred to indicate a faster seepage path from the lake. Failure initiated by internal erosion within the upper tephra (Fig. 10), following the appearance of seepage from the toe of erosion scarp E1

Table 2 Failure sequence of the tephra barrier on 18 March 2007

Time of photograph (NZST = UTC+12 hrs) 06:30 09:54 09:55 to 09:57 09:57 to 10:05 10:45 11:07 to 11:08 11:09 to 11:10

11:11 to 11:15 11:14 11:17 to 11:20 11:21 11:22 11:22 to 11:36/11:38 11:45 to 11:57 12:02 a

Eventa Rainstorm following period of prolonged wet weather caused lake level to rise (2535.7 to 2535.8 AMSL) Significant seepage apparent from toe of erosion scarps E1 and C0 (Fig. 9a) The eastern flank of erosion scarp E1 started to retrogress towards erosion scarp E3, creating a new narrow (5 to 7 m wide) erosion scarp. Erosion scarps E3, W1, W4, W5 and W6 remained unchanged. (Fig. 9a). ERLAWS Geophone 1, started to respond at 09:56 Apparent tension cracks form immediately above the western flank and head of erosion Scarp E4. Spike on Geophone 1 at 10:05, could be related to the development of these tension cracks and possible retrogression of E4 Spike doublet on Geophone 1 could be related to retrogression of E4 Rapid headward erosion and enlarging of the new erosion scarp extending from the western flank of E1 (Fig. 9b). Largest spike on Geophone 1 at 11:07 Rapid enlarging and retrogression of the new erosion scarp into E1 and E3 (and E4) to form one large channel, extending towards the lake. Water can be seen escaping from the toe of this feature in the location of C0 and E1. GNS lake data (derived from the bubbler sensor), indicates some drawdown of the lake (3 cm in 8 min), starting at 11:10 (DOC sensor not sensitive enough to detect). These data indicate an initial breach of the dam along the eastern flank and water flow along the newly formed channel (Fig. 9b) The western flank of the channel continues to retrogress and enlarge, towards erosion scarp W1 (Fig. 9c) Significant flow of water noted from the toe of the channel (at the toe of erosion scarp E1) The western flank of the channel continues to erode and retrogress towards W1, in a series of relatively small failures off the channel flank. (Fig. 9c) Rapid large failure of the downstream dam face, forming one large failure scarp, extending from W6 in the east towards the initial dam-breach channel inlet in the west. Seepage noted from the toe of the failure scarp (Fig. 8d) Rapid failure of remaining dam, leading to large-scale breach upstream of erosion scarps W6, W4, W1, C0 and E1. Flow of water from the lake begins through a wide channel. ERLAWS tripwire pulls out at 11:22. Lake starts to draw down rapidly (Fig. 9) Channel enlarges as the side walls collapse. Peak discharge of 543 m3 /s recorded at 11:26. (Fig. 9). 11:24 to 11:32 broad peak recorded on Geophone 1 Series of peaks recorded on Geophone 1, possibly related to breach widening as the flanks collapsed Breach fully widened to 45 m

Note that all of the instruments were recording off independent internal clocks set approximately (apparently within about a minute) to New Zealand Standard Time

Landslides 7 • (2010)

311

Recent Landslides a)

CRATER

b)

LAKE

N

CRATER

LAKE

N

Initial breach 11:10 am Lake level at failure 2535.8 m

Lake level 11:10 at failure 2535.8 m 9 :0

11

11:08 11:07

W6

W4

9:57 9:56

E1

W6

W1

E4

ICE

10 m

c)

Dam camera

CRATER

5

.10

C0

E3

9:5

544 9.

544 9.

0

11

5

9:5

C0

9:57 9:56

E1

W4

W1

0

d)

LAKE

E3

E4

ICE

10 m

9 :0 11

Dam camera

CRATER

LAKE

N

N

Main breach 11:22 am

Initial breach 11:10 am 11:36

Initial breach 11:10 am

Lake level 11:10 at failure 2535.8 m

11:10

:222 11:2

9

:0

11

W1

11 C0

ICE

11:16

11

5 9:5

11 .10

CO

E3 E4

Dam camera

0

10 m

9 :0

11

9:57 9:56

E1

W1

54

10 m

:20

W4

11:07

11:18

9.

544 9.

0

.10

W6

11:08

11:11 11:13 11:14 11:17

09

: 11

9:57 9:56

E1

2

1

W4

:2

11:36 11:36

W6

0 11:18 1:2

9

:0

11

:211 11:2

11

11:08 11:07

11:11 11:13 11:14 11:17

Lake level at failure 2535.8 m

5

9:5

E3 E4

ICE

Dam camera

KEY 1-m contours derived from laser scanning carried out 4/3/2007

E4

Approximate extent of ice taken from photographs on 4/3/2007

Head of erosion scarps recorded 4/3/2007 Visible water flow/seepage Seepage location (evidence of sub-surface water flow)

Lake level at failure

11:10 11:08

Location of failure scarp with time of first appearance Approximate location of failure scarp with time of first appearance

Fig. 9 Barrier-failure sequence derived from the monitoring equipment installed on the barrier

at 09:54 (Fig. 9). The time-lapse photographs indicate that erosion continued retrogressing upslope into the barrier towards the lake, forming a narrow channel, which eventually reached to the lake (Fig. 9). This led to an initial surface breach at 11:10 (NZST), and a minor lowering of the lake-level of 0.03 m in 8 min as recorded by the GNS bubble-in level sensor (Manville et al. 2007a, b). Between 11:15 and 11:20 (NZST), flow of water through the breach at an average rate of 5 m3/s appeared to initiate undercutting and slumping along its western flank (right-hand, 312

Landslides 7 • (2010)

looking downstream). The eastern flank of the growing channel was outside the camera field-of-view. The removal of toe support appeared to initiate further instability of the downstream slope, and a series of large, very rapid slope failures occurred between 11:20 and 11:21, which retrogressed towards the crest of the barrier. The final, large breach of the dam occurred at 11:22 when retrogressive slope failures led to loss of toe support and failure of the remaining barrier against the hydrostatic pressure of the lake water. This created a breach approximately 35 m wide, which

A

A’

Retrogressing erosion scarps

North

Lake level at failure 2535.8

Pre- failure ground surface (4 Mar 07)

11

:10

11

:09

11 :08 :07

11

South

9:5 9: 7 56

Crown of erosion scarp E1 8 Feb 07 and 4 Mar 07 Seepage point E1 hot (2531.7) Ice 18 Jan 07

Lake level post failure 2529.4

Ice 8 Feb

Post failure ground surface

Ice 4 Mar Ice 23 Mar

B

B’

North

18 MAR 07 Lake level at failure 2535.8

South

11:21 Rapid large-scale retrogressive failure 11:22 of dam Main breach

11:20 Rapid failure of downstream slope along erosion scarps C0 and E1

11:11 - 11:18 Erosion of toe of C0 from water flow associated with initial breach C0 Seepage point (2529.9) Ice 18 Jan 07

Lake level post failure 2529.4

Ice 8 Feb 07

Post failure ground surface 10 m

(V=H)

Ice 4 Mar 07 Ice 23 Mar 07

Fig. 10 Cross-sections summarising the failure sequence (refer to Fig. 5 for the diagram key)

enlarged to 45 m by 12:02 that extended down to the top of the lava, resulting in a peak outflow of about 530 m3/s (Manville et al. 2007a, b). Observations and measurements made over the days and weeks following the failure indicated that headward erosion, which created the initial breach at 11:10 (NZST), initiated within the upper tephra, creating a narrow (15-m wide) terrace, with the level of the erosion terrace corresponding to the base of the fine tephra overlying the lower lapilli layer. Geomorphic mapping of the barrier after failure indicated that this terrace had been subsequently incised along its

western flank by large outflows associated with the main barrier failure at 11:22 (NZST). The escaping Crater Lake water entrained large volumes of snow and ice, talus, older lahar terrace deposits, colluvium, and landslide debris as it flowed down the steep convoluted channel of the upper Whangaehu River, bulking up by a factor of about 3.4 and transforming into a hyper-concentrated flow (Pierson 2005). This included approximately 106 m3 of material from the toe of a relict landslide, now known as the Whangaehu landslide (Fig. 11), located

Whangaehu River (lahar travel path)

Fig. 11 Aerial oblique view of the Whangaehu landslide

Relict Headscarp

Crater Lake

Headscarp of 2007reactivation

2007 Debris

100m

Landslides 7 • (2010)

313

Recent Landslides along the southern flank of the channel. The loss of toe support caused the landslide to reactivate and dam the river at several points. These dams subsequently breached forming multiple debris pulses (Massey et al. 2007) in the channel. Discussion The response to lake filling and failure mechanisms of the tephra barrier derived from detailed assessment of information obtained leading up to and during failure of the barrier were similar to those identified in the initial report by Hancox et al. (1997), and refined later by STAP (2003) with one almost incidental exception. The conclusions of STAP (2003) and Gillon et al. (2006), were: 1. The barrier was unlikely to breach at low lake levels (<2,531 m); 2. The barrier was almost certain to breach when the lake approached or exceeded 2,536 m; 3. At lower lake levels, the major contributors to likely failure where assessed as earthquake-induced liquefaction, or surge waves caused by rockfall and tephra sliding into the Crater Lake. 4. At higher lake levels, slope instability was assessed as the major contributor, followed by overtopping (by wind-induced waves), internal erosion, phreatic eruption wave, or surges caused by a tephra slide or calving from the ice cliff on the northwestern side of the lake. The range of possible scenarios did not include increased seepage due to direct infiltration of heavy storm rainfall into the tephra dam, which added to the already high ground water level. Pore water reducing particle-to-particle frictional resistance had been discussed, but was considered to be unlikely due to the tephra grain-size characteristics and overburden pressure. A more direct role of simply adding to the increasing seepage volume was not considered. The direct infiltration may have hastened the initiation of failure, but possibly by hours rather than days, because nearly similar rainfall intensities earlier in the week, when the lake was only slightly lower, had not initiated failure or enough erosion to be visible in photographs taken before 09:54 on 18 March. Although not without initial “teething problems” with false alarms, the ERLAWS system proved to be robust and reliable, and its essential components survived the March 2007 lake break-out and remain functional to this day although the response alert level has dropped back to “normal” (Table 1). The remaining instruments detected subsequent debris flows (lahars) in the channel (triggered by a sudden phreatic eruption in the lake on September 2007), but these were too small to trigger any warning signal, and none was required. Conclusions The failure sequence of a tephra dam impounding Ruapehu’s Crater Lake occurred when the lake was about 1.1 m below its crest (overtopping level 2,536.9 m). Failure was triggered by the rising lake and increased seepage forces, induced by a period of wet weather (13 to 15 March 2007) and a rainstorm on morning of 18 March 2007. The proximal failure mechanism was headward erosion by retrogressive slumping, caused by internal seepage and erosion, initiating at seepage point E1-hot, which formed the most rapid seepage path within the dam. Headward erosion was confined to the upper tephra and lapilli layers, which sapped through the dam leading to an initial breach of the barrier at 11:10. This in turn triggered undercutting of the downstream slope of the barrier, and initiated large, rapid (11:20 314

Landslides 7 • (2010)

to 11:22 NZST) slope instability, which led to the main breach of the barrier and peak outflows of about 530 m3/s (Manville et al. 2007a, b). Continuing investigations are aimed at assessing the geotechnical properties of the dam, along with finite-element groundwater and stability analyses. The break-out lahar itself is the subject of a variety of ongoing studies. The ERLAWS instruments and plans remain in place to warn of the ongoing hazard of future lahars. Acknowledgements This work was funded by a variety of agencies of the New Zealand Government. The authors acknowledge the following people for their input to this investigation: Robin Fell (Unisearch Ltd), Murray Gillon (DamWatch), Neville Palmer (GNS Science), Stuart Read (GNS Science) and Reece Gardener (Astrolabe). The paper benefited from helpful comments by Professor Chia-Nan Liu and an anonymous reviewer. References Bornas MA, Team QR (2002) The 2001 Mount Pinatubo Crater Lake breakout crisis, 23rd Annual PNOC-EDC Geothermal Conference, Hotel Inter-Continental. Makati City, Philippines, pp 7–16 Bornas MA, Tungol N, Maximo RPR, Paladio-Melosantos ML, Mirabueno HT, Javier DV, Corpuz EG, Dela Cruz EG, Ramos AF, Marilla JD, Villacorte EU (2003) Caldera-rim breach and lahar from Mt. Pinatubo, Philippines: Natural breaching and resulting lahar. Proceedings, IUGG 2003 Congress, Sapporo, Japan, p A558 Carrivick JL, Manville V, Cronin SJ (2009) A fluid dynamics approach to modelling the 18th March 2007 lahar at Mt. Ruapehu, New Zealand. Bull Volcanol 71:153– 169 Christenson BW, Wood CP (1993) Evolution of a vent-hosted hydrothermal system beneath Ruapehu Crater Lake, New Zealand. Bull Volcanol 55:547–565 Cronin SJ, Hodgson KA, Neall VE, Palmer AS, Lecointre JA (1997a) 1995 Ruapehu lahars in relation to the late Holocene lahars of Whangaehu River, New Zealand. N Z J Geol Geophys 40(4):507–520 Cronin SJ, Neall VE, Lecointre JA, Palmer AS (1997b) Changes in Whangaehu River lahar characteristics during the 1995 eruption sequence, Ruapehu volcano, New Zealand. J Volcanol Geotherm Res 76:47–61 Gillon M, Fell R, Keys HJR, Foster M (2006) A contribution to public safety studies— failure modes and likelihood analysis for a tephra dam on an active volcano. Bulletin of ANCOLD (Australian National Committee on Large Dams), 134. pp. 11–22 Hancox G, Nairn I, Otway P, Webby M, Perrin N, Keys H (1997) Stability assessment of Mt Ruapehu crater rim following the 1995–1996 eruptions. GNS client report 43605B Hancox G, Keys H, Webby M (2001) Assessment and mitigation of dam-break lahar hazards from Mt Ruapehu Crater Lake following the 1995–96 eruptions. In: Geotechnical Society 2001 symposium, Engineering & Development in Hazardous Terrain. Christchurch, 24–25 August 2001 Hodgson KA, Manville V (1999) Sedimentology and flow behaviour of a rain-triggered lahar, Mangatoetoenui Stream, Ruapehu volcano, New Zealand. Geol Soc Amer Bull 111:743–754 Hodgson KA, Lecointre J, Neall VE (2007) Onetapu formation: the last 2000 yr of laharic activity at Ruapehu volcano, New Zealand. N Z J Geol Geophys 50(2):81–99 Johnston DM, Houghton BF, Neall VE, Ronan KR, Paton D (2000) Impacts of the 1945 and 1995–1996 Ruapehu eruptions, New Zealand: an example of increasing societal vulnerability. Geol Soc Amer bull 112:720–726 Keys H (2003) Minor change confirmed in the elevation of the lava surface controlling the overflow of Crater Lake, Mount Ruapehu. Department of Conservation internal report for the Crater Lake Scientific and Technical Advisory Panel Keys HJR, Green PM (2008) Ruapehu lahar New Zealand 18 March 2007: Lessons for hazard assessment and risk mitigation 1995–2007. Journal of Disaster Research 3 (4):284–296 Kilgour GN, Manville V, Della Pasqua F, Graettinger A, Hodgson KA, Jolly G (2009) The 25 September 2007 eruption of Mount Ruapehu, New Zealand: directed ballistics, surtseyan jets, and ice-slurry lahars. J Volcanol Geotherm Res in press Lecointre J, Hodgson KA, Neall VE, Cronin SJ (2004) Lahar-triggering mechanisms and hazard at Ruapehu volcano, New Zealand. Nat Hazards 31:85–109 Lube G, Cronin SJ, Procter JN (2009) Explaining the extreme mobility of volcanic iceslurry flows, Ruapehu volcano, New Zealand. Geology 37:15–18

Manville V (2004) Palaeohydraulic analysis of the 1953 Tangiwai lahar: New Zealand’s worst volcanic disaster. Acta Vulcanologica XVI(1/2):137–152 Manville V, Cronin SJ (2007) Break-out lahar from New Zealand’s Crater Lake. EOS Trans AGU 88:441–442 Manville V, Hodgson KA, Houghton BF, Keys HJR, White JDL (2000) Tephra, snow and water: complex sedimentary responses at an active, snow-capped stratovolcano, Ruapehu, New Zealand. Bull Volcanol 62:278–293 Manville V, Hodgson KA, Nairn IA (2007a) A review of break-out floods from volcanogenic lakes in New Zealand. N Z J Geol Geophys 52:131–150 Manville V, Massey C, Hancox G, Keys H (2007b) Characterising the initiation of the 18th March 2007 Ruapehu Crater Lake Lahar. IUGG Special session of the IAVCEI Commission of volcanogenic sediments VS022. Perugia, Italy Massey CI, Nelis S, Keys HJR (2007) Whangaehu Landslide—Preliminary findings from the inspection carried out on 25April 2007. GeoNet landslide response report Nairn IA, Wood CP, Hewson CAY (1979) Phreatic eruptions of Ruapehu: April 1975. N Z J Geol Geophys 22:155–173 New Zealand Geotechnical Society (2005) Description of soil and rock. Guidelines for the field classification and description of soil and rock for engineering purposes O’Shea BE (1954) Ruapehu and the Tangiwai disaster. N Z J Sci Technol B36:174–189 Pierson TC (2005) Hyperconcentrated flow—transitional process between water flow and debris flow. In: Jakob M, Hungr O (eds) Debris-flow hazards and related phenomena. Praxis. Springer, Berlin, pp 159–202 STAP (2003) Second report to Minister on possible hazard created by filling of Crater Lake, Ruapehu and potential lahar. Crater Lake Scientific and Technical Advisory Panel, 14 May 2002

Turner G, Ingham M, Bibby H (2007) Electrical resistivity monitoring of seepage and stability of the tephra barrier at Crater Lake, Mt Ruapehu, New Zealand. Geophys Res Abstr 9:11630

Vernon Manville is formerly from GNS Science. C. I. Massey ()) . G. H. Hancox . M. McSaveney GNS Science, Avalon, New Zealand e-mail: [email protected] V. H. Manville J. Keys · C. Lawrence University DepartmentofofLeeds, Conservation, Leeds, Kingdom Turangi,United New Zealand Present address: V. Manville Lawrence H. J. Keysof. C. University Leeds, Department Conservation, Leeds, UnitedofKingdom Turangi, New Zealand Present address: C. Lawrence C. NZ Lawrence Forest Managers, NZ ForestNew Managers, Turangi, Zealand Turangi, New Zealand

Landslides 7 • (2010)

315