Patterns of movement in reactivated landslides

    Patterns of movement in reactivated landslides C.I. Massey, D.N. Petley, M.J. McSaveney PII: DOI: Reference:

S0013-7952(13)00089-6 doi: 10.1016/j.enggeo.2013.03.011 ENGEO 3562

To appear in:

Engineering Geology

Received date: Revised date: Accepted date:

4 October 2012 24 January 2013 2 March 2013

Please cite this article as: Massey, C.I., Petley, D.N., McSaveney, M.J., Patterns of movement in reactivated landslides, Engineering Geology (2013), doi: 10.1016/j.enggeo.2013.03.011

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Patterns of movement in reactivated landslides 1

2

1

C.I. Massey , D.N. Petley , M.J. McSaveney 1

T

The Institute of Geological and Nuclear Science, Avalon, Lower Hutt, New Zealand

2

IP

The University of Durham, UK.

SC R

Corresponding author: C.I.Massey. [email protected]. Tele: +64 (0) 4570 4770

ABSTRACT

The primary aim of this research was to study the relationship between landslide motion and

6

NU

its causes, with reference to large, slow moving, reactivated translational rock slides. Surface 3

displacements of the 2210 m Utiku landslide, in central North Island, New Zealand were

MA

measured using continuous GPS (cGPS), for three years. The nature of the movement of such slides has often been difficult to determine because of poor temporal and spatial monitoring resolutions. After removal of tectonic plate motion, the temporal pattern of the

TE

D

landslide’s surface motion could be understood to arise from irregular episodes of faster (upto-21 mm/day) and slower (up to 26 mm/year) post-failure landslide displacement, and

CE P

seasonal cyclic displacements of about 20 mm/yr – 10 mm per half year in alternating directions. Intervals of faster motion gave rise to displacements of between 10 to 120 mm per event. Faster displacement was associated mostly with basal sliding (mechanism 1), involving

AC

deformation within a thin clay seam as recorded by borehole inclinometer surveys. Slower surface displacement involved permanent internal deformation of the larger landslide mass, consisting of plastic deformation within the landslide body and / or slip along existing internal planes of weakness, and slip on the slide base (mechanism 2); it accounted for up to 26 mm/year of displacement at a mean angle of about 49 from the horizontal, indicating that the slide mass was thinning as it moved down slope. Seasonal cyclic displacements were synchronous with changes in pore pressure, suggesting that it is a shrink/swell process (mechanism 3) associated with wetting and recharge of groundwater during the wetter winter months, leading to downslope movement, and soil shrinkage leading to up slope rebounds during the dryer summer months. The brief periods of faster displacement were triggered by seasonal peaks in pore pressure, linked to long periods (12 to 20 weeks) of increased precipitation and lowered evapotranspiration. Faster displacement, however, was not arrested

1

ACCEPTED MANUSCRIPT by lowering pore pressure or by any other monitored factor. Similarly, periods of slower displacement did not correlate with pore-pressure changes, or with any other monitored factor. This study has shown that the annual movement pattern of a reactivated landslide is a

T

combination of these processes that generate a complex overall movement record. The field

IP

measurements showed real variability arising from variations in rainfall and pore pressure,

1.

SC R

which were overprinted with measurement noise that may mask some other processes.

INTRODUCTION

NU

The movement of reactivated landslides can impose a large financial cost on society. Lee and Jones (2004) and Mansour et al. (2011) note that whilst reactivated landslides generally do

MA

not kill many people, they are responsible for high levels of economic loss. For example, the 1956-9 Portuguese Bend landslide in California (which caused no fatalities) resulted in losses and court-imposed damages of US$86 million ($680 million equivalent value in 2013). In

D

many parts of the world, unexpected reactivation of landslides generates the highest levels of

TE

non-coastal landslide hazard (Lee and Jones, 2004). Understanding the behaviour of reactivated landslides is therefore important, but at present detailed knowledge of these

CE P

processes is surprisingly scant, primarily because the datasets that can be used for such

AC

analyses have been difficult to collect.

A range of techniques have become available in recent years that allow monitoring of patterns of landslide movement with improved temporal and spatial resolution. Application of these techniques, which include laser-based geodetic techniques, differential GPS, and ground- and satellite-based radar, when combined with monitoring of potential causal and triggering factors, is providing new insights into the mechanisms of slope deformation (Thiebes, 2012). In turn, this permits improved slope management, and is paving the way for better forecasting and prediction of likely landslide behaviour (Petley, 2010; Thiebes, 2012).

In order to link landslide-causal factors to their consequences, in this case movement patterns, high-precision measurement of the movement of reactivated landslides over a representative time period (several years rather than days or months) is required, coupled

2

ACCEPTED MANUSCRIPT with analyses of the resulting datasets. Whilst some studies address this issue (e.g. Corominas, 2005; van Asch, 2007; Petley et al., 2005 and Thiebes, 2012), the development of approaches for such analyses are in general lagging behind the advances in the ability to

T

measure and create the datasets. It is only through detailed analyses of a wide range of

SC R

IP

landslide types in diverse materials that general principles of movement will be understood.

For specific landslides under study, Corominas et al., 2005, Gonzalez et al. (2008), and Matsuura et al. (2008), found that velocity increased non-linearly as pore pressure increased.

NU

In addition, Bertini et al., (1984) and Gonzalez et al., (2008) showed that for the same value of pore pressure, the velocity when groundwater was rising was higher than during lowering

MA

(Picarelli, 2007), implying that the relationship between shear stress and normal stress, as assumed by the Mohr-Coulomb failure criterion may in some cases be non-linear. The relationship between pore pressure (groundwater) and landslide movement can be

D

complicated by complex landslide geology and hydrogeology, in particular by the contrasting

TE

permeability of intact, fissured and sheared materials forming the slide mass, the occurrence of multiple slip surfaces within the slide mass leading to a complex displacement profile with

CE P

depth, and the presence of large-scale heterogeneities providing direct conduits for surface water into the landslide (e.g. Corominas et al., 1999; van Asch, et al., 2007, 2009).

AC

Consequently, large, slow slides often show an erratic and complex response to hydrological input (Corominas, 2005; Malet; et al., 2005; van Asch et al., 2007, 2009).

It has been hypothesised that landslide velocity, although clearly linked to pore-pressureinduced changes in effective stress, is also governed by rate-induced changes in shear strength of the materials, caused by behaviour of the clay particles during shearing (Lupini et al., 1981; Skempton, 1985; Angeli et al., 1996; Picarelli, 2007); and/or consolidation and strength regain during periods of rest (Nieuwenhuis, 1991; Angeli et al., 2004). It has also been proposed that shear-strength parameters, represented as c’ and ’ in the Mohr-Coulomb failure criterion, can be modified by inclusion of a viscous resistance component (Bertini et al., 1984; Leroueil et al., 1996; Corominas et al., 2005; van Asch, 2007, 2009; Picarelli, 2007; Gonzalez et al., 2008). Many authors have used viscosity functions to better describe and in

3

ACCEPTED MANUSCRIPT some cases predict the motion patterns of these types of landslide assuming that once motion is triggered, the landslides move by visco-plastic flow, rather than by rigid-plastic frictional slip, (e.g. Iverson 1985; Angeli, et al. 1996; Corominas et al. 2005; van Asch et al. 2008;

IP

T

Ranalli et al. 2009.

SC R

This paper proposes a framework for understanding the movements in reactivated landslides, based upon a high-quality dataset from one such landslide, located at Utiku in the North Island of New Zealand (Fig 1). The Utiku landslide is a large, reactivated, deep-seated,

NU

translational landslide that displays comparatively low rates of movement. Nonetheless, the landslide movement has repeatedly damaged both a major highway and an important railway

MA

that traverse the landslide. This study reports on three years of high-resolution monitoring, and explores relationships between the patterns of movement and their controlling factors. Even though the geometry of the landslide is comparatively simple, and the materials

D

controlling failure are not complex, the observed patterns of movement are not

TE

straightforward. Novel techniques for movement analysis allow the underlying patterns to be determined. Based upon this analysis, a framework through which the movement of

The Utiku landslide

AC

1.1

CE P

reactivated landslides can be understood is proposed.

The Utiku landslide is located at 39.75S, 175.83E in the central part of North Island, New Zealand (Fig. 1). The climate is temperate oceanic in the Köppen-Geiger classification, and characterised by warm summers (December through to February) where the average daily temperature is 22C, and cooler winters (June to August), where the average daily temperature is 11C. Rainfall does not vary significantly between winter (mean monthly rainfall 70 mm), and summer (mean monthly rainfall 81 mm), with a mean annual rainfall of about 960 mm.

The landslide volume is about 22 10 m , and it can be classified as a reactivated, deep6

3

seated, translational landslide in the Cruden and Varnes (1996) scheme. Prior to this study, the landslide had been monitored at a low temporal resolution since 1965; measurements 4

ACCEPTED MANUSCRIPT during this period indicate that it moved extremely slowly to very slowly (16 mm/yr < x < 1.6 m/yr) (Stout, 1977). Nonetheless, both the North Island Main Trunk railway line (NIMT) and State Highway 1 (SH1), which cross the landslide, have been repeatedly damaged by

IP

T

movements, although they remain in use (Fig. 2).

SC R

Although the Utiku landslide is interesting in itself, its observed patterns of movement broadly illustrate the behaviour of large translational mass movements in weakly consolidated Neogene materials, primarily silty, fine sandstones and sandy siltstones of marine origin, in

NU

New Zealand and other sites of similar materials around the world. In New Zealand there are 2

over 7,000 mapped landslides each with a plan area >10,000 m in these materials. Some of

MA

these are shown on Fig. 3. The majority of these are slow-moving, relatively deep-seated, translational landslides (Dellow et al., 2005; Massey, 2010). The large number of landslides in these Neogene deposits results from a combination of factors, including the low intact

D

strength of the slide mass (typical unconfined compressive strengths are in the order of <1 to

TE

20 MPa (Read and Miller, 1990)); deeply incised rivers; extensive tectonic folding, and faulting; bedding-plane defects (bedding-plane shears); and regionally persistent, bedding-

CE P

parallel clay seams along which sliding may occur (Stout, 1977; Thompson, 1982; Mountjoy

AC

and Pettinga, 2006; Reyes, 2007).

The regional tectonic setting of the Utiku landslide is one of gentle folding from east-west compression (Lee et al., 2012). Associated uplift rates are about 1.5 to 2.0 mm/yr (Pillans, 1986; Pullford and Stern, 2004), which have driven fluvial incision. The regional dip of bedding is 3° – 7° SSE (Lee et al., 2012).

The Utiku landslide has formed in sediments that lie at the stratigraphic boundary between the early to mid-Pliocene, Tarare Sandstone and Taihape Mudstone (Lee et al., 2012). Both formations are described in the engineering-geological classification of New Zealand Geotechnical Society (2005) as extremely weak to weak, blue-grey (when fresh) or yellowishbrown (when highly weathered), very silty sandstones, with flattened (ovoid to irregular) calcareous concretions (average diameter 0.2 m). Bedding in both materials includes the

5

ACCEPTED MANUSCRIPT presence of thin (2-20 mm) clay seams of predominantly smectite (Thompson, 1982; Reyes, 2007), at about eight different stratigraphic levels (Stout, 1977; Thompson, 1982). The layers

T

represent marine-deposited ash from large eruptions from North Island volcanos.

2

IP

The landslide has a plan area of about 800,000 m , and consists of two main areas (Fig. 3). 2

SC R

The lower area, of about 260,000 m , has been active in the last 50 years. It is about 400 m wide and 1,100 m long, and extends southward from a fresh headscarp that lies close to State Highway 1 (SH1), downslope to the actively eroding toe at the Hautapu River. The

NU

second area is inactive and is located upslope from the active area; there are no records that this area has moved significantly in at least the last 150 years. This paper relates only to the

MA

active area of the landslide.

The main geological materials found in the landslide area are, in reverse chronological order

D

(i.e. youngest first): landslide debris, further subdivided into three sub-units – intact displaced

TE

blocks of sandstone (landslide crest), partially remoulded rafts of intact sandstone and remoulded sandstone (landslide toe), representing different proportions of remoulding during

CE P

movement, which increases with distance and displacement from the landslide crest; landslide slip surface clay; river-terrace gravels; the in-situ Tarare sandstone; and the in-situ

AC

Taihape mudstone. The assumed slip surface is formed within a thin clay layer which varies in thickness from 0.05 to 0.2 m (Table 1).

2.

METHODS

The Utiku landslide was investigated through field mapping, the drilling of six boreholes through the basal shear surface, analyses of historical movements and the analysis of measurements from borehole piezometers, inclinometers and a rain gauge, all of which were installed in July 2008 (for locations, see Fig. 3). Surface movements on the active landslide have been measured since 1965 using a network of pegs, surveyed using theodolites, but at a rather variable frequency (for locations, see Fig. 3).

A network capable of high temporal and spatial resolution monitoring of landslide movement

6

ACCEPTED MANUSCRIPT and of the variables that influence movement, was installed on the landslide in 2008. Monitoring equipment was selected primarily for the temporal resolution that could be

Installed monitoring networks

IP

2.1

T

achieved, so that periods of landslide movement could be linked to the triggering factor(s).

SC R

Since July 2008, surface movements have been measured by four continuous GPS (cGPS) stations located on the landslide, referenced against two cGPS stations located on stable terrain off the landslide (station THAP is located 6 km to the north of the landslide and station

NU

UTKU, 0.5 km east of the landslide). Full details of the monitoring system are contained in Massey (2010). The displacements were analysed using Bernese v5.0 software holding IGS

MA

final orbits and Earth orientation parameters fixed, following the methodology of Wallace and Beavan (2010). Data from the landslide stations were corrected using the two external sites to remove the regional tectonic signal; a continental scale velocity of about 10 mm/yr SSW

TE

D

relative to a fixed Australian plate (Wallace and Beavan, 2004).

The cGPS monitoring was exceptionally reliable, with data being collected for 99.4% of the

CE P

study period for each instrument on average.

The root mean square error on the daily

position (at 1, 68% confidence) of the cGPS stations was ±6.0 mm E-W, ±5.4 mm N-S and

AC

±18 mm vertically.

Inclinometer tubes were installed in two boreholes (BH1A and BH3A on Fig. 3). Measurements were made at approximately three-monthly intervals, or when significant surface movement of the landslide had been detected. Inclinometer accuracy is ±6 mm over 25 m of tubing (Slope Indicator, 2005) and the measurement precision between any two surveys is about ±2 mm (at 1).

Pore pressures within the landslide were measured using Casagrande piezometer standpipes in boreholes PZA and BH1 to BH4 (Fig. 3). Each piezometer tube measured piezometric pressure within a screened and bentonite-sealed response zone, with two tubes measuring two response zones per hole. The depth and length of each response zone was selected after

7

ACCEPTED MANUSCRIPT examination of the borehole logs. The lower response zones corresponded to the logged landslide slip surface, into which vibrating-wire pressure transducers were installed. Other

T

standpipes were monitored with a manual dip meter.

IP

All pore-pressure measurements used in these analyses were recorded at the tip of the

SC R

vibrating-wire sensor. Hourly readings (in Hertz) from vibrating-wire piezometers BH1, BH2, BH3, BH4 and PZA were converted to pressure and averaged over each 24-hour period (UTC), to obtain a daily averaged pore pressure for each instrument. Readings were

NU

corrected for barometric effects by subtracting the daily mean change in barometric pressure, using data from a barometric pressure sensor installed on the landslide. The stated accuracy

MA

of the vibrating-wire piezometers is ±0.1% of the operating pressure (Geokon, 2005), which for the Utiku piezometers was 700 kPa, giving an error of ±0.7 kPa (about ±0.07 m of piezometric head).

The instruments were less reliable than for the displacement

D

measurements and resulted in some data gaps. Nonetheless, there were enough data for

TE

analysis of the role of pore pressure.

CE P

Rainfall on the landslide was recorded by a tipping-bucket rain gauge (0.2 mm per tip) located at site UTK4 (Fig. 3). The 0.2 mm increments recorded over each 24-hour period (midnight to

3.

AC

midnight UTC) were summed to determine daily rainfalls.

STRUCTURE OF THE ACTIVE PART OF THE UTIKU LANDSLIDE

The active (lower) area of the Utiku landslide can be further divided into two broad zones: an upper zone comprising displaced, relatively intact blocks (rafts) of Tarare Sandstone. These transition into a lower zone consisting of predominantly remoulded materials, with numerous closely-spaced tension cracks and landslide scarps. On this basis, it may be classified as a complex, reactivated, translational rockslide-earthflow (Cruden and Varnes, 1996) or blockslide (Panet, 1969).

The structure of the landslide is, however, unusual. The main headscarp of the active portion – the upper limit of the 1964 reactivation – is a near-vertical scarp up to 6 m high. Since 1991,

8

ACCEPTED MANUSCRIPT the road upslope of this scarp has subsided, indicating retrogression. The western flank of the active landslide is marked by a steep (45° – 60°) linear lateral scarp that is stepped and which ranges in height from about 20 m in the north to 60 m in the south. This feature is more

T

distinct towards the southern edge of the landslide, where Toe Toe Road crosses its upper

IP

part (Fig. 2). Subsidence of Toe Toe Road outside of the scarp was noted in September 2010

SC R

indicating that this flank of the landslide has also enlarged. The eastern flank is not so well defined, but is marked by exposure or near-exposure of the slip surface, where landslide

NU

debris is missing or thin.

The surface of the upper active part of the landslide has several large hillocks up to 40 m

MA

high, with side slopes of 35° to 45°. A tension crack approximately 3 m deep, 2 m wide and 100 m long extends up the flank and along the crest of one hillock (Fig 3). Minor landslide scarps are apparent on its flanks, particularly on the downslope (eastern face), indicating that

D

the block is breaking up. Between the hillocks are small valleys, interpreted as grabens

TE

between slide-block horsts. Natural drainage lines (recorded on 1950’s aerial photography,

grabens.

CE P

but now modified by various historical minor mitigation works to drain the landslide) follow the

AC

The toe of the landslide, adjacent to the Hautapu River, is a highly active zone that can be classified as an earthflow (Cruden and Varnes, 1996). In this area, the landslide materials are largely remoulded but contain some small rafts of sandstone. There is surface ponding of water, but slopes are about 10°. Upslope (to the north), the surface is broken by closelyspaced (typically < 1 m), fresh tension cracks, with vertical offsets of 0.5–2.0 m. The Hautapu River is eroding the toe of the landslide, although none of the historical movement periods can be attributed to river erosion.

Six boreholes drilled into the landslide in 2008 (Table 1) identified a laterally persistent clay layer that has been interpreted as the primary shear surface at the base of the landslide (Fig. 4). It comprises dark grey, soft, silty clay (highly plastic) of the smectite group, with minor angular fine gravel formed from fine-grained sandstone clasts, and with a well-developed

9

ACCEPTED MANUSCRIPT shear fabric (slickensides). The fine to coarse gravel-sized clasts are found in the upper part of the clay and are interpreted as comminuted Tarare Sandstone. The thickness of the shear zone (clay layer and comminuted sandstone) varies from 0.05 to 0.2 m. The slip surface

T

corresponds to the uppermost and thickest of the bedding-plane clay layers within the

IP

Taihape Mudstone. Three-point solutions between different combinations of borehole (e.g.

SC R

BH1, BH2 and BH3, and between BH2, BH3 and BH4) and historical observations from the slip surface exposed in dewatering shafts excavated through it in 1969, indicate the slip surface is planar, with a dip/dip direction of 7°/230°. Field mapping identified a clay layer: in

NU

the highway cut batter slope (Fig. 3); at the toe of the landslide close to the Hautapu River level; and in the western cliff of the Hautapu River. The dip/dip direction of bedding was

MA

typically 3–7°/230–240°, coincident with the dip/dip direction of the landslide slip surface. Structural contours of the slip surface are shown on Fig. 3.

D

The role of this clay layer in controlling movement has been investigated using the

TE

inclinometers in boreholes BH1A and BH3A. Cumulative displacements from the inclinometer tubes indicate the presence of a single slip surface at depths corresponding to the clay

CE P

surfaces identified in drillhole logs (Fig. 5 and Table 2). The inclinometers confirmed a relatively thin (< 0.5 m, where 0.5 m is the measurement increment) slide surface that was

AC

almost horizontal (Fig. 4), when projected along the main landslide movement bearing and taking into account the dip of the slip-surface clay seam. The BH1A inclinometer data also showed sinusoidal deformation between 12 m and 26 m below ground level, but this probably represents buckling of the inclinometer casing (Stark and Choi, 2008) into voids between the casing and the in-situ ground. Borehole logs for these depths corroborate the presence of voids. Thus, multiple strands of evidence suggest that movement has occurred on one bedding-parallel clay horizon. Geotechnical tests of the clay from the shear surface were undertaken using ring-shear apparatus (Massey, 2010; Kilsby, 2007). The material appears to be in a residual strength state with laboratory test results indicating a cohesion of 4 (±6) kPa and a friction angle of 8.3 (±1)° (uncertainties 1).

4.

MONITORING RESULTS

10

ACCEPTED MANUSCRIPT 4.1

Movement patterns of the Utiku landslide

Displacements of the landslide over the monitoring period were well constrained. The data indicate that movement was primarily as blocks sliding almost along the strike of the bedding-

T

parallel clay layer (compare the orientation of the vectors in Fig. 6 with the structure contours

IP

in Fig. 3), such that movement is not down-dip. In the upper portion of the landslide, the net

SC R

movement vector has a bearing of about 140°. Downslope, in the lower landslide, the movement vector was rotated slightly to a bearing of about 155°. Tension cracks on the landslide body reflected these movement directions (Fig. 6). Thus, in both cases the

NU

movement was, perhaps surprisingly, sub-parallel to strike. In the upper portion of the active landslide, this gives an apparent dip in the direction of movement on the shear surface of

MA

about 1°, whereas in the lower portion of the landslide the apparent dip is about 3°. The landslide was therefore a wedge that thickened towards the west. The lateral (western) release plane was formed by the persistent and stepped western scarp (approximate dip/dip

D

direction 60°/065°), which constrained the landslide from moving directly down dip. The top of

TE

the wedge corresponded to the landslide head scarp (approximate dip/dip direction 60°/110°), which was also sub-parallel to the trend of grabens. The plunge and trend of the wedge

CE P

intersection was approximately 2°/154° and was inside the landslide movement envelope

AC

(Fig. 3).

Cumulative displacement plots of selected survey marks (Fig. 7A) showed that historical motions were not constant over time, but were stepped, with the steps being associated with periods of more rapid displacement. These steps were particularly frequent during 1965 – 1972, when the landslide was in its most active monitored phase. This period generally corresponded with prolonged wet weather in the area (Fig. 7B). However, between 1973 and 1978, a positive cumulative rainfall deviation occurred but without any discernible step in the cumulative displacement plots. No surface-movement data exists for this period on the lower landslide.

These data suggest a complex relationship between precipitation, pore pressure and landslide movement. This relationship has been further complicated by mitigation measures,

11

ACCEPTED MANUSCRIPT comprising surface drainage and two pump shafts, which were installed in 1969. It was not possible to quantify the effects of these measures in this study: the pumps stopped working in 1976, and the surface drainage was not maintained and rapidly deteriorated. Relationships

T

between antecedent rainfall and pore pressure are discussed in Section 4.2, using the results

SC R

IP

from the recent monitoring.

The cGPS data indicated significant horizontal displacements between June 2008 and February 2011 (Fig. 8). Station velocities (and thus landslide movements) were non-linear in

NU

time. To assist in identifying patterns in station velocity, the filtered time series have been plotted as cumulative horizontal displacement along the main movement bearings (Fig. 9). In

MA

the western part of the active landslide, cGPS station UTK1 on the lower landslide showed the largest movement, while Stations UTK3 and UTK4 suggest decreasing displacement towards the head scarp (Fig. 9 and Table 3). On the eastern part of the active landslide,

D

station UTK2 showed a slightly larger displacement compared to UTK4. The flatter

TE

morphology of the area around UTK2 and the more remoulded and disturbed nature of the core samples from BH2 in comparison to those from BH4, suggests that this is a different

CE P

slide block within the larger landslide and that this area had undergone larger displacement. These surface displacement magnitudes and vectors (Fig. 6) suggest that displacement

AC

increases from the crest to the toe of the landslide, and also laterally across the landslide from west to east. These patterns also indicate rotation of the landslide mass about the western landslide scarp (Massey 2010).

The horizontal cGPS time-series data (Fig. 9) showed three main types of motion, all of which could be classified as “slow” in the classification of Cruden and Varnes 1996:

1) Short periods of faster displacement (steep positive gradients, or steps, in cumulative horizontal displacement plots). These occurred over periods of a few days to a few weeks, and were responsible for most of the recorded displacement; 2) Longer periods of slower motion, comprising semi-constant displacement rates lasting many months to years, termed here “slow displacement”; and

12

ACCEPTED MANUSCRIPT 3) Slow, cyclic motion which appeared to follow a seasonal cycle. These movement types are described in more detail below.

Faster displacement

T

4.1.1

IP

Four periods of faster displacement were recorded (Table 4 and Fig. 9). Within the daily

SC R

temporal resolution of the data, the timing of the onset of faster-displacement periods 1, 3 and 4 were synchronous across the stations. Faster displacement period 2 was different in that it was only recorded by UTK1 on the lower landslide. During faster displacements, motion was

NU

largely horizontal with a translational angle of displacement (downslope from the horizontal)

MA

closely comparable to the 1 slope of the basal slip surface in that direction.

For each faster displacement period, peak speeds were reached four to six days after initiation of acceleration; followed by an equally rapid deceleration over a similar period until

D

the faster displacement ended. Faster displacement period 4 was an exception where the rate

TE

of displacement decreased over about nine days. The patterns were also shown in station

CE P

acceleration (Fig. 10); stations accelerated rapidly from rest, and achieved peak accelerations within two to three days, followed by equally rapid deceleration.

AC

Acceleration of station UTK1 during faster displacement period 2 showed a “saw-tooth” pattern, with periods of acceleration followed by periods of deceleration with a similar duration and magnitude, over eight to ten days. These patterns were recurrent throughout the period. This suggested that this movement period actually comprised numerous short-duration, smallmagnitude events, each similar to those shown during faster displacement periods 1, 3 and 4. This pattern of displacement was similar to the stick-slip displacements observed by Allison and Brunsden (1990) and in the laboratory by Ng and Petley (2009).

4.1.2

Slower displacement

Slower displacement occurred at very slow rates that were semi-constant in both horizontal and vertical components. Slower displacement probably occurred throughout the monitoring period, and was inferred to occur, but be undetectable, during periods of faster displacement.

13

ACCEPTED MANUSCRIPT

Patterns of horizontal slow displacement were examined by deleting faster displacements and their intervals from the cumulative horizontal-time series from each station. Lines were fitted

T

to the residual time series; the gradients of these lines were used to estimate the slow

IP

displacement rate (Table 5, Fig. 11). Horizontal slower displacement increases from the head

SC R

scarp to the toe, with cGPS UTK3 (the station nearest the head scarp), moving at 2.5 (0.1) mm/yr (uncertainties at 1, 65% confidence), and UTK1 (nearest the toe) moving at 22.8 (0.2) mm/yr (Fig. 11). Differences in the rate of slow displacement appeared to correspond

NU

to the changing material properties of the landslide debris as it transitioned from the landslide head scarp – where it comprised relatively intact blocks of sandstone – through the landslide

MA

to the toe – where it comprised remoulded sandstone. In addition, the orientation of the displacement vector in the lower landslide indicated that the landslide was moving more parallel to the dip of the slip surface – resulting in a steeper slip surface – rather than along its

TE

D

strike as shown in the upper landslide.

CE P

Vertical motions were assessed by fitting lines to the vertical component trends of the cGPS time series, and the line gradient was used to estimate the vertical movement rate for each station (Table 5, Fig. 11). Negative (downward) vertical motions were statistically significant

AC

for all stations, ranging from 5 to 77 mm/yr. The time series for UTK2, UTK3 and UTK4 showed linear rates over the entire interval (Fig. 11B-D). However, UTK1 showed an initial faster rate for the first six months with downward motion of 77 (4) mm/yr, followed by a period of slower motion, 2.6 (0.7) mm/yr, which continued at a constant rate for the remainder of the interval (Fig. 11A).

Gradients in the horizontal and vertical motion time series were used to calculate translational displacement angles during periods of slower displacement. These angles ranged between 18 and 61 from the horizontal (Fig. 11), and differed significantly from the 1–3 apparent slope of the basal slide surface. The steepest angle of translation (61) was recorded at cGPS station UTK3, located in a graben and could in part be related to the downward motion of the graben slide block. However, UTK1, UTK2 and UTK3 are located on horst features, up 14

ACCEPTED MANUSCRIPT slope of any scarps, and no rotation is evident in plots of displacement or in the mature trees on the surface.

Seasonal cyclic displacement

T

4.1.3

IP

Each time series has been further detrended by removing the slow displacement component.

SC R

The residuals showed statistically significant cyclic displacement (Fig. 12). For the horizontal series, the residuals from UTK1, 2 and 4 show a repeating pattern of up- and down-slope changes in horizontal motion with about ±10 mm/yr amplitude (i.e. 5 mm per half year)

NU

through the summer and winter, suggesting a seasonal origin. Upslope motion is represented by increasing displacement and downslope motion by decreasing displacement (Fig. 12).

MA

Downslope motion occurred during winter and upslope motion in summer. The de-trended time series from UTK2 and 3 were nearly identical (peak to peak), while peaks in the UTK1

D

time series lagged those of UTK2 and 3 by about two months.

TE

Seasonal cycles were only able to be resolved in the vertical time series for UTK3 (Fig. 12), where annual cycles represented movements of about 20 mm/yr (10 mm per half year in

CE P

alternating directions) with upward movement in winter. The peak-to-peak vertical pattern correlated with those from the UTK2 and 3 horizontal time-series. Seasonal cycles in the

AC

vertical time series from the other stations, if present, were too small to be resolved.

It is not known to what depth of the landslide is affected by this seasonal movement. No evidence of seasonal cycles has been determined in inclinometer measurements of displacement at the slide base. It can be assumed that the seasonal displacements affect the landslide to at least the depth of the bases of the cGPS monuments – steel I-beams grouted into the landslide mass to depths of between 4 and 5 m below ground level.

4.1.4

Inclinometer displacements

Whilst inclinometer-derived subsurface motions have a low temporal resolution, the magnitude and timing of horizontal deformation between the slip surface and ground surface has been assessed using inclinometer BH3A and cGPS station UTK3, which are located on

15

ACCEPTED MANUSCRIPT the landslide surface about 5 m apart. The data showed that the patterns of displacement between the slip surface and the ground surface for the same period were consistent, with periods of faster displacement at the surface corresponding to periods of more rapid

IP

SC R

slip surface were also consistent within measurement precision.

T

displacement at the slip surface, and that the bearings of motion at the ground surface and

Displacements were calculated for the time period corresponding to the first and last inclinometer measurements (9/07/2008 to 13/05/2010, about 673 days), before it was

NU

sheared off sometime between the 15/05/2010 and 2/10/2010 surveys (Table 6). For the 9/07/2008 – 13/05/2010 period, the data indicated that surface displacement was about

MA

double the displacement recorded at the slip surface, with the majority of the difference occurring during the longer, slower displacement periods between the periods of faster

D

displacement.

TE

Inclinometer displacements during the slower displacement periods were further examined by comparing ground-surface and slip-surface displacements for the period 10/11/2008 to

CE P

13/05/2010 (about 549 days) during which no period of faster displacement occurred. The data indicate a mean horizontal basal sliding speed of 2 (±1) mm/yr, and a mean horizontal speed of

15 (±4) mm/yr (uncertainties 1). These data indicated that surface

AC

surface

displacement was significantly greater than displacement recorded at the landslide basal slip surface. Similar results have been reported by Massey (2010) for the Taihape landslide in New Zealand, but in that case the inclinometer surveys were more frequent and over a longer time period.

4.2

Rainfall, groundwater and landslide movement

The highest basal pore-pressures were recorded in the upper part of the active landslide, where the basal slip surface is comparatively deeper. Downslope, the landslide thickness decreases (Fig. 4A) and pore pressures measured above the slip surface also decrease. All piezometers showed seasonal cycles, with pore pressure decreasing in summer and autumn months (December to May) and rising in winter and spring months (June to November) (Fig. 16

ACCEPTED MANUSCRIPT 13). Winter groundwater levels were close to or at the ground surface in the upper part of the landslide (piezometers BH4 and BH3) (Fig. 4) throughout the observation period.

T

Histograms of daily changes in pore pressure for all piezometers (Fig. 14A-C) were unimodal,

IP

and were approximately normally distributed with a range of ±3 kPa, and a standard deviation

SC R

of about ±1 kPa. Pore-pressure gradients for both the rising and recessional limbs of the timeseries data recorded at BH3, BH4 and PZA, were about 0.1 kPa/day, and for BH1, about 0.2

NU

kPa/day.

Although pore-pressure changes appear to be linked to rainfall, histograms of pore-pressure

MA

change suggested that pore pressures responded only slowly to rainfall. Correlation between piezometer pore pressures and the antecedent rainfall on the landslide, may establish the time frame over which rainfall influenced piezometer response, thereby establishing the

D

period of antecedent rainfall that is needed to cause a statistically significant change in daily

TE

pore pressure.

CE P

The pore pressure at a given time and the accumulated antecedent rainfall were assessed incrementally, using daily mean pore pressures and daily rainfall totals. The correlation

AC

analysis assumes a linear relationship between pore pressure and rainfall. Correlation was obtained by dividing the covariance of the two variables by the product of their standard deviations. Results (Fig. 15) showed that the correlation was highest for antecedent rainfalls accumulated over 12 to 20 weeks, indicating that pore pressures were responding to long periods of antecedent rainfall. These results implied that pore-pressure responses, triggering the periods of faster displacement resulted from long periods of wet weather rather than large magnitude, short-duration rainfall.

Histograms of deviations from mean pore pressure for all piezometers were seasonally bimodal (Fig. 14D-F) between summer and winter (Fig. 13). Deviations from mean pore pressures indicate that mean seasonal fluctuations in piezometric head were about 18.7 kPa (1.9 m) above and 12.8 kPa (1.3 m) below mean values, representing seasonal fluctuations of

17

ACCEPTED MANUSCRIPT between 3% and 6%. Piezometer BH1 showed the largest seasonal variation, and PZA, the lowest.

T

The relationship between surface displacement rate and associated pore pressure at the

IP

landslide base have been investigated by comparing the cGPS and piezometer data from

SC R

stations UTK1 and UTK3, located 15 and 10 m horizontal distance from PZA and BH3 respectively. Fig. 16 shows that the periods of faster displacement coincided with seasonal peaks in pore pressure, linked to long periods (12 to 20 weeks) of increased precipitation and

NU

lowered evapotranspiration. However, periods of slower displacement appeared to be

MA

independent of pore pressure.

Although periods of faster displacement were initiated by seasonal pore-pressure peaks, their cessation was independent of lowering pore pressure or any other monitored factor. Graphs

D

of pore pressure versus cumulative displacement (Fig. 17A), and pore pressure versus

TE

displacement rate (Fig. 17B) for faster displacement period 4 showed an initial rapid increase in displacement rate to peak values as pore pressure increased. However, as post-peak

CE P

displacement rates decreased, pore pressure either remained constant, or in some cases actually increased. This was also the case for faster displacement periods 1 to 3 where

AC

cessation of motion did not coincide with a decrease in pore pressure. This indicated that the landslide could be moving faster or slower at the same value of pore pressure, and that the relationship between pore pressure and faster displacement was complex. This complexity is discussed in more detail in Section 5.

Seasonal cyclic displacement has been investigated by comparing pore pressure and detrended cumulative displacement for PZA and UTK1, and BH3 and UTK3 (Fig. 18). The seasonal cycles in both pore pressure and displacement rate were in phase and had similar relative amplitudes, suggesting that groundwater was somehow controlling this behaviour.

5.

SYNTHESIS AND MECHANISMS OF UTIKU LANDSLIDE MOVEMENT PATTERNS

18

ACCEPTED MANUSCRIPT The movement of the Utiku landslide could be described as episodic post-failure landslide displacement. In the classification of Leroueil et al. (1996), the movement patterns can be classified as stages three (reactivate occasionally) and four (active), as movement is

T

occurring along a fully developed slip surface, which is at residual strength (Skempton, 1985).

IP

The head scarp of the Utiku landslide was actively retrogressing, but the debris has yet to be

SC R

fully incorporated into the landslide mass; in these areas, the movement of the landslide may be classified as stages 1 and 2 (pre- and post-failure) (Leroueil et al., 1996).

NU

Utiku landslide motion patterns were repetitive, with three types of slow motion predominating: faster displacement, slower displacement and seasonal cyclic displacement,

MA

punctuated by periods of rest. The displacements vary in both magnitude and duration resulting in unsteady, non-uniform motion.

D

These motion patterns are common to other slow, very slow and extremely slow, reactivated

TE

landslides (Cruden and Varnes 1996). Long-term monitoring of such landslides (e.g. Angeli et al., 1996; Carey, 2011; Schulz et al., 2009a; Zangerl et al., 2010) suggest that these patterns

CE P

can alternate for many years, and that the status quo is repeating periods of faster movement of varying magnitudes and rates, interrupting longer periods of slower displacement, seasonal

AC

cyclic displacement and / or inactivity.

We infer three deformation mechanisms for the surface and sub-surface motion of the Utiku landslide: 1) basal sliding; 2) plastic deformation; and 3) seasonal shrinkage and swelling. These three mechanisms appeared to account for the measured patterns of movement.

5.1

Basal sliding

Frictional basal sliding was inferred to be the mechanism for the simultaneous periods of faster displacement of the ground surface and on the slip surface. This motion accounts for the largest recorded displacements, both before and during the period of high temporalresolution monitoring. Displacement vectors, and longitudinal angles of translation, were predominantly horizontal, and essentially parallel to the apparent-dip angle of the slip surface. 19

ACCEPTED MANUSCRIPT

Examples from monitored similar landslides from the literature show that there is a general relationship between pore pressure and landslide velocity. For example, detailed observations

T

of slopes sliding on pre-existing slip surfaces at San Martino in Italy (Bertini, 1984) and at

IP

Salledes in France (Cartier and Pouget, 1988) show that displacement rate follows variations

SC R

in the pore-water-pressure regime, where the rate of movement depends on the applied shear stress (Leroueil et al., 1996). However, previous research has found that plots of pore pressure versus displacement rate show hysteresis, indicating that different landslide

NU

velocities can occur at the same pore pressure (Nakamura, 1984; van Asch et al., 2007; Gonzalez et al., 2008; Matsuura et al., 2008; Bertini, 1984), depending on whether pore

MA

pressure is rising or falling.

At the Utiku landslide, the initial relationship between increasing pore pressure and increasing

D

surface velocity is linear, and similar to other published relationships (e.g. Nakamura, 1984,

TE

van Asch et al., 2007; and Gonzalez et al., 2008). Landslide velocity therefore might be expected to increase with pore-pressure induced decreases in the shear resistance of the slip

CE P

surface and vice versa as observed by Bertini et al. (1984) and Cartier and Pouget (1998). However, at Utiku, velocity tends not to slow as pore pressure reduces, even though the latter

AC

should induce increased shear resistance.

The Utiku landslide behaviour suggested that once slip occurred, there was no relationship between pore-pressure-induced change in the residual shear resistance of the slip surface and landslide velocity or that the relationship is weak or highly non-linear. Previous research has used viscosity functions to better describe and in some cases predict the motion patterns of these types of landslide assuming that once motion is triggered, the landslides move as visco-plastic flows, rather than rigid-plastic frictional slip (e.g. Iverson 1985; Angeli et al. 1996; Corominas et al. 2005; van Asch et al. 2008; and Ranalli et al.2009).

Gonzalez et al. (2008) found that a constant viscosity gives a linear relationship between pore pressure and displacement rate. At Utiku, in common with other landslides showing

20

ACCEPTED MANUSCRIPT hysteresis (Bertini et al. 1984; Gonzalez et al. 2008; Matsuura et al. 2008), the relationship between pore pressure and velocity is non-linear suggesting that apparent viscosity and

T

therefore apparent shear resistance were controlled by other factor(s).

IP

These other factors may, for example, include: rate-induced changes in shear strength of the

SC R

slip-surface clay (i.e. a dynamic rather than static frictional resistance), caused by either rearranging the clay-particle bonds during shearing; and/or consolidation of the clay during motion. However, increasing resistance forces are also likely to occur as a result of

NU

progressive changes in landslide geometry. For example, the movement direction of the Utiku landslide should cause increasing shear resistance along its western flank (Massey, 2010).

MA

The interaction between the various slide-blocks forming the landslide mass, for example buttressing and mass transfer, are factors that Morgenstern (1995) and Ferrari et al., (2010),

Plastic deformation of the landslide mass

TE

5.2

D

identified as important controls on slow-moving landslides.

CE P

Slow displacement accounted for mean displacements of about 17 mm/yr at angles of about 39 from the horizontal. The angle of the movement vector associated with slower displacement was steeper than the angle of the basal slip surface, which suggested that this

AC

movement was not primarily associated with basal sliding. The greater dip angle suggested that the mechanism of displacement comprised a combination of localised downward motion of slide-blocks in grabens, and wider plastic deformation of the landslide mass, which increases in magnitude towards the toe as the debris becomes more disaggregated. The landslide mass is therefore not rigid, but is thinning as it moves downslope. These mechanisms may in-part drive displacement on the slide surface, as some basal sliding during periods of slower displacement were recorded. This slower component of surface motion appears to be unrelated to pore-pressure fluctuations, because, within the limits of detection, the rate of slow displacement did not vary with changing pore pressure, or with any other monitored factor.

5.3

Seasonal shrinkage and swelling 21

ACCEPTED MANUSCRIPT The seasonal cyclic displacements appeared to be exactly synchronous with changes in pore pressure, implying that they may be a shrink/swell process associated with wetting and recharge of groundwater during the winter months, leading to slope-parallel lengthening and

T

soil shrinkage leading to slope-parallel shortening during the dryer summer months. These

IP

patterns may also be associated with vertical movement of the ground surface in response to

SC R

shrink/swell, and we note that vertical motion is much less well resolved from the cGPS time series than is horizontal motion. This mechanism may be similar to that reported by Leung

6.

NU

and Ng (2012).

A FRAMEWORK FOR INTERPRETING MOVEMENT PATTERNS OF SLOWLY

MA

MOVING REACTIVATED LANDSLIDES

Based on the data from the movement of the Utiku landslide, a framework for interpreting the

D

movement patterns of reactivated landslides in similar settings and materials is as follows.

TE

When reactivated, movement in such landslides appears to consist of the following key mechanisms:

CE P

1. Cumulative basal sliding;

2. Cumulative permanent internal deformation, consisting of:

AC

a. Plastic deformation within the landslide body and / or b. Slip along existing internal planes of weakness

3. Recoverable (reversible) internal deformation, primarily as a result of shrink/swell processes.

The basal sliding mechanisms are modulated primarily by factors that alter the stress state of the landslide. In most cases (e.g. Angeli et al. 1996; Leroueil et al. 1996; Moore et al. 2007; Schulz et al. 2009b; and Zangerl et al. 2010), these consist of either an increase in pore pressure or, more rarely, seismic disturbance, toe excavation or head loading and even atmospheric tides. Given that pore-water pressure change is the most common driver of reactivated movement, and that precipitation and evapotranspiration are often seasonal, the resulting pattern of movement may show seasonality. In many cases, slow movement can

22

ACCEPTED MANUSCRIPT continue through both wet and dry seasons, but during the wet season, sliding movement can vary in rate with shorter periods of faster movement (Fig. 19).

T

There is also non-recoverable internal cumulative deformation within the landslide. In the

IP

Utiku landslide most of the recorded movement of this type can be explained as internal

SC R

plastic deformation of the landslide debris, which drives a very low rate of thinning of the landslide mass and results in a downhill movement vector. In some landslides, this might also cause more rapid movement events due to pore-pressure induced flow processes within the

NU

landslide mass. However, where there is local compression, the landslide mass may thicken,

MA

and the surface rise relative to the basal surface.

Finally, there will generally be a seasonal movement of the landslide surface, with a very small magnitude, associated with wetting and drying of the landslide mass. This may be

D

detected only as a small upslope and downslope component of movement, but may be driven

TE

by vertical movement, where this component in large enough to be detected. Seasonal cyclic

annual cycle.

CE P

displacement may also drive a small net downslope displacement over the course of the

AC

The annual movement of a reactivated landslide is a combination of these processes, which combine to generate a complex overall movement record. Field measurements will contain natural variability from variations in rainfall and pore pressures (as discussed), overprinted with measurement noise, which may mask some other processes.

7.

CONCLUSIONS

The Utiku landslide-displacement time series has revealed patterns of episodic post-failure slow displacement comprising periods of faster and slower displacement and seasonal cyclic slow displacement, punctuated by intervals of rest, all of which occur with varying frequencies and durations through the seasons.

Intervals of faster displacement gave rise to displacements of between 10 to 120 mm per

23

ACCEPTED MANUSCRIPT event, at rates of up-to-21 mm/day. Faster displacement was mostly from basal sliding, involving frictional slip within a thin clay seam as recorded by borehole inclinometer surveys. The periods of faster displacement were triggered by seasonal peaks in pore pressure, linked

T

to long periods (12 to 18 weeks) of rain and lowered evapotranspiration. Faster displacement,

SC R

IP

however, was not slowed by lowering pore pressure or by any other monitored factor.

Slower displacement involved slip on the slide base, plastic deformation within the slide mass and / or the localised downward motion of slide-blocks in grabens. It accounted for mean

NU

displacements of about 17 mm/yr at a mean angle of about 39, indicating that the slide mass thins as it moves down slope. As with faster displacements, periods of slower displacement

MA

did not correlate with pore-pressure changes, or with any other monitored factor.

Seasonal cyclic slow displacements were synchronous with changes in pore pressures,

D

implying that it was a shrink/swell process associated with wetting and recharge of

TE

groundwater during the wetter winter months (leading to apparent downslope movement) and

CE P

soil shrinkage during the dryer summer months (leading to up slope rebounds). These surface motions were not detectable along the basal slip surface.

AC

Based on the data from the movement of the Utiku Landslide, a framework for interpreting the movement of reactivated landslides in similar settings and materials has been proposed. Movement in such landslides may consist of the following key mechanisms: 1) basal sliding; 2) permanent internal deformation, consisting of plastic deformation and / or slip along existing internal planes of weakness, within the landslide body; and 3) recoverable (reversible) internal deformation, primarily as a result of shrink/swell processes.

ACKNOWLEDGMENTS The authors would like to acknowledge: the New Zealand Natural Hazard Platform for funding this research; the GeoNet project for funding the monitoring equipment and the GeoNet team for installing and maintaining the equipment; and Kiwi Rail for funding the subsurface geotechnical investigations. The authors would also like to thank the two anonymous

24

ACCEPTED MANUSCRIPT reviewers for their helpful comments.

T

REFERENCES

IP

Allison, R.J., Brunsden, D. 1990. Some mudslide movement patterns. Earth Surface

SC R

Processes and Landforms 15 (4) pp. 297 - 311.

Anderson, J. M., Mikhail, E. M. 1998. Surveying theory and practice, seventh edition. WCB McGraw-Hill. pp. 50-71.

NU

Angeli, M. G., Gasparetto, P. Menotti, R. M., Pasuto, H., Silvano, S. 1996. A visco-plastic model for slope analysis applied to a mud slide in Cortina d’Ampezzo, Italy. Quarterly

MA

Journal of Engineering Geology. Vol. 29. pp. 233 – 240.

Angeli, M. G., Gasparetto, P., Bromhead, E. 2004. Strength-regain mechanisms in

D

intermittently moving landslides. In: Proceedings of the 9th International Symposium on

TE

Landslides. Rio de Janeiro. Vol. 1. Taylor and Francis, London (pub). pp. 689 – 696.

CE P

Bertini, T., Cugusi, F., D’Elia, B., Rossi-Doria, M. 1984. Climatic conditions and slow movements of colluvial covers in Central Italy. In: Proceedings of the IV International

AC

Symposium on landslides, Toronto 1984. Vol 1. pp 367 – 376.

Carey, J. 2011. The progressive development and post-failure behaviour of deep-seated landslide complexes. Doctoral thesis, Durham University, UK. Cartier, G., Pouget, P. 1988. Etude du compartement d’un remblai construit sur un versant instable. Le remblai de Salledes (Puy de Dome). Report No. 153. Laboratoire Central de Ponts et Chaussee Paris. France.

Corominas, J., Moya, J., Ledesma, A., Lloret, A., Gili, J. A. 1999. Monitoring of the Vallcebre landslide, Eastern Pyrenees, Spain. In: Proceedings of the International symposium on Slope Stability Engineering. IS-Shikoku’ 99. Matsuyama. Japan. pp 1239 - 1244.

Corominas, J., Moya, J., Ledesma, A., Lloret, A., Gili, J. A. 2005. Prediction of ground displacements and velocities from groundwater level changes at the Vallcebre landslide 25

ACCEPTED MANUSCRIPT (Eastern Pyrenees, Spain). In: Landslides. Vol. 2. pp. 83-96.

Cruden, D.M., Varnes, D.J., 1996. Landslide types and processes, in: Turner, K.A., Schuster,

T

R.L. (Eds.), Landslides: investigation and mitigation. Special Report 247, Transportation

IP

Research Board, Washington, pp. 36–75.

SC R

Dellow, G.D., McSaveney, M.J., Stirling, M.W., Berryman, K.R., 2005. A probabilistic landslide hazard model for New Zealand, in: Pettinga, J.R.; Wandres, A.M. (Eds.) Geological Society of New Zealand 50th annual conference, 28 November to 1 December 2005,

NU

Kaikoura: programme & abstracts.. Geological Society of New Zealand miscellaneous publication 119A, p. 24.

MA

Ferrari, A., Ledesma, A., Gonzalez, D. A., and Corominas, J. 2010. Effects of the foot evolution on the behaviour of slow-moving landslides. In: Engineering Geology. DOI:

2005.

Model

4500

TE

Geokon,

D

10.1016/j.enggeo.2010.11.001.

vibrating

wire

piezometer.

Data

sheet.

CE P

http://www.geokon.com/products/datasheets/4500.pdf.

Gonzalez, D. A., Ledesma, A., Corominas, J. 2008. The viscous component in slow moving landslides: A practical case. In: Landslides and engineering slopes. Chen et al. (eds). pp

AC

237 – 242.

Iverson, R.M., 1985. Dynamics of slow landslides: a theory of time dependent behaviour. In: Abrahams, A. D. (Ed), Hillslope processes. Binghampton Symposium in Geomorphology No. 16. pp. 297 – 317.

Kilsby, C. 2007. An engineering geological appraisal of the Utiku Landslide, North Island, New Zealand. M.Sc. Thesis, University of Portsmouth. pp. 19 – 27.

Lee, E.M. and Jones, D.K.C. 2004. Landslide Risk Assessment. Thomas Telford, London.

Lee, J.M., Bland, K.J., Townsend, D.B., Kamp, P.J.J. (comp). 2012. Geology of the Hawke's Bay area. Institute of Geological and Nuclear Sciences 1:250 000 geological map 8. Lower Hutt. Institute of Geological and Nuclear Sciences Limited. 26

ACCEPTED MANUSCRIPT Leroueil, S., Locat, J., Vaunat, J., Picarelli, L. and Faure, R. 1996. Geotechnical characterisation of slope movements. In. Proceedings of the Seventh International Symposium on Landslides, (Ed. K. Senneset) Trondheim, Norway, Balkema, Rotterdam.

IP

T

1, pp.53-74.

Leung, A.K., Ng, C.W.W. 2012. Seasonal movement and groundwater flow mechanism in an

SC R

unsaturated saprolitic hillslope. In: Landslides Journal. DOI 10.1007/s10346-012-0343-7.

Lupini, J. F., Skinner, A. E., Vaughn, P. R. 1981. The drained residual strength of cohesive

NU

soils. In: Geotechnique. Vol. 31. No. 2. pp. 181 – 213.

Malet, J. P., Van Asch, Th. W. J., Van Beek, R., Maquaire, O. 2005. Forecasting the

MA

behaviour of complex landslides with a spatially distributed hydrological model. In: Natural Hazards and Earth System Sciences.

D

Mansour, M.F., Morgenstern, N.R., Martin, D., 2011. Expected damage from displacement of

TE

slow-moving slides. Landslides 8(1), 117–131. DOI 10.1007/s10346-010-0227-7.

CE P

Massey, C.I. 2010. The dynamics of reactivated landslides: Utiku and Taihape, North Island, New Zealand. PhD Thesis, Durham University.

AC

Matsuura, S., Asano, S., Okamoto, T. 2008. Relationship between rain and/or meltwater, pore-water pressure and displacement of a reactivated landslide. In: Engineering Geology. Vol. 101. pp. 49 – 59.

Moore, R., Carey, J., McInnes, R.G. & Houghton, J. 2007. Climate change , so what? Implications for ground movement and landslide frequency in Ventnor Undercliff, Isle of White. In: McInnes, R.G., Jakeways, J., Fairbank, H. & Mathie, E. (eds) Landslides and Climate Change: Challenges and Solutions. Balkema, Rotterdam, 335–344.

Morgenstern, N.R. 1995. Keynote paper: The role of analysis in the evaluation of slope stability. Landslides, Bell (ed). Balkema, Rotterdam.

Mountjoy, J., Pettinga, J. R., 2006. Controls on large deep-seated landslides in soft rock terrain. Rock mass defects and seismic triggering, in: Earthquakes and Urban 27

ACCEPTED MANUSCRIPT Developments. Proceedings of the New Zealand Geotechnical Society 2006 Symposium, Proceedings of Technical Groups. 31, IPENZ.

T

Nakamura, H. 1984. Landslides in silts and sands mainly in Japan. In: Proceedings of the IV

IP

International Symposium on landslides, Toronto 1984. Vol 1. pp 155 – 185.

SC R

Nieuwenhuis, J. D. 1991. Variations in the stability and displacements of a shallow seasonal landslide in varved clays.

New Zealand Geotechnical Society, 2005. Description of soil and rock: Guidelines for the field

NU

classification and description of soil and rock for engineering purposes. New Zealand

MA

Geotechnical Society, 38 pp.

Ng, K-Y. and Petley, D.N. 2009. A process approach towards landslide risk management in Hong Kong. Quarterly Journal of Engineering Geology and Hydrogeology 42(4), pp. 487-

TE

D

498.

Panet, M., 1969. Discussion of K.W. John’s Paper (ASCE Proc. Paper 5865, March 1968).

CE P

Journal of the Soil Mechanics and Foundation Division, ASCE 95(SM2), pp 685–686.

Petley, D.N. 2010. Landslide Hazards. Alcántara-Ayala, I. and Goudie, A. (eds)

74.

AC

Geomorphological Hazards and Disaster Prevention. Cambridge University Press. pp 63-

Petley, D.N., Mantovani, F., Bulmer, M.H.K.., and Zannoni, F. 2005. The interpretation of landslide monitoring data for movement forecasting. Geomorphology, 66(1-4), pp.133147.

Picarelli, L. 2007. Considerations about the mechanics of slow active landslides in clay. In: Progress in landslide science. Sassa, K., Fukuoka, H. Wang, F., Wang, G. editors. Chapter 3. pp. 27-57.

Pillans, B., 1986. A Late Quaternary uplift map for North Island, New Zealand. Royal Society of New Zealand, Bulletin 24. pp. 409–417.

28

ACCEPTED MANUSCRIPT Pulford, A., Stern, T., 2004. Pliocene exhumation and landscape evolution of central North Island, New Zealand: The role of the upper mantle. Journal of Geophysical Research 109,

T

F01016, doi:10.1029/2003JF000046.

IP

Ranalli, M., Gottardi, G., Medina-Cetina, Z., Nadim, F. 2009. Uncertainty quantification in the

Online first. DOI 10.1007/s10346-009-0185-0.

SC R

calibration of a dynamic viscoplastic model of slow slope movements. In. Landslides.

Read, S.A.L., Miller, P.J. 1990. Characterisation and classification of New Zealand Tertiary

NU

age sedimentary soft rocks. In. Bridge design and research seminar, 1990. Volume 4: Road engineering in soft rock materials. RRU Bulletin 84. pp. 1 to 25.Reyes, A.G., 2007.

MA

Petrological study of selected Taihape landslide rock samples, in: The Institute of Geological and Nuclear Sciences Science Report 2007/32.

D

Schulz, W. H., McKenna, J. P., Kibler, J. D., Biavati, G. 2009a. Relations between hydrology

TE

and velocity of a continuously moving landslide – evidence of pore=pressure feedback

CE P

regulating landslide motion? In. Landslides. Vol. 6. pp 181-190.

Schulz, W. H., Kean, J. W., Wang, G. 2009b. Landslide movement in southwest Colorado triggered by atmospheric tides. In. Nature Geoscience. 2. pp. 863-866.

AC

Skempton, A. W. 1985. Residual strength of clays in landslide, folded strata and the laboratory. In: Geotechnique. Vol. 35. No. 1. pp. 3 – 18.

Slope Indicator, 2005. Digitilt inclinometer probe. Data sheet. Durham Geo Slope Indicator. http://www.slopeindicator.com/pdf/digitilt-vertical-inclinometer-probe-datasheet.pdf

Stark, T.D. and H. Choi, "Slope Inclinometers for Landslides," Landslides, Journal of the International Consortium on Landslides, 5(3), pp. 339-50.Stout, M.L., 1977. The Utiku landslide, North Island, New Zealand. Geological Society of America, Reviews in Engineering Geology III, pp 171–184.

Stout, M. L. 1977. The Utiku landslide, North Island, New Zealand. In: Geological Society of America, Reviews in Engineering Geology. III. pp 171 to 184.

29

ACCEPTED MANUSCRIPT Thiebes, B. 2012. Landslide analysis and early warning systems. Local and regional case study in the Swabian Alb, Germany. Doctoral thesis accepted by the University of Vienna,

T

Austria.

IP

Thompson, R.C. 1982. Relationship of geology to slope failures in soft rocks of the Taihape-

SC R

Mangweka area, Central North Island, New Zealand. PhD Thesis, University of Auckland.

Van Asch, W. J., Van Beek, L. P. H., Bogaard. 2007. Problems in predicting the mobility of slow-moving landslides. In: Engineering Geology. Vol 91. pp 46 – 55.

NU

Van Asch, W. J., Malet J. P., Bogaard, T. A. 2009. The effect of groundwater fluctuations on the velocity pattern of slow-moving landslides. Nat. Hazards Earth Syst. Sci., 9, 739–749,

MA

2009.

Van Genuchten, P. M. B., De Rijke, H. 1988. On pore pressure variations causing slide

D

velocities and accelerations in seasonally active landslide. In: Earth Surface Processes

TE

and Landforms. Vol. 14. Issue 6. pp 577-586.

North

CE P

Wallace, L. M., Beavan, J. 2004. Subduction zone coupling and tectonic block rotations in the Island,

New

Zealand.

Journal

of

Geophysical

Research.

109.

doi:

AC

10.1029/2004JB003241.

Wallace, L. M., Beavan, J. 2010, Diverse slow slip behaviour at the Hikurangi subduction margin, New Zealand. J. Geophys. Res., 115, B12402. doi:10.1029/2010JB007717.

Zangerl, C. Eberhardt, E. Perzlmaier, S. 2010. Kinematic behaviour and velocity characteristics of a complex deep-seated crystalline rockslide system in relation to its interaction with a dam reservoir. Engineering Geology. 112. pp. 53-67.

30

ACCEPTED MANUSCRIPT Table 1. Landslide slip surface details derived from boreholes. Borehole locations are shown in Fig. 3. Thickness (m)

Description

BH1

65.15

313.96

0.05

BH1A

65.82

313.38

0.10

Silty CLAY, minor angular fine gravel of fine grained sandstone; dark grey, soft. With shear fabric.

15.42

331.48

0.08

BH3

49.05

302.72

0.05

BH3A

48.90

302.90

0.10

BH4

28.05

302.55

0.20

Silty CLAY, some angular fine gravel of fine grained sandstone; dark grey to black, soft. Brecciated with shear fabric.

SC R

BH2

T

Slip surface (m RL)

CLAY, minor angular fine to coarse gravel of fine grained sandstone; dark grey, soft tending firm to stiff with depth. With shear fabric. CLAY; dark grey speckled light grey, firm. With shear fabric.

NU

1

1

Depth below ground level

IP

Borehole

Material descriptions as per New Zealand Geotechnical Society (2005)

(m RL) 379.7

BH3A

352.3

Depth of logged slip 1 surface

Total displacement 2 in period

Movement 3 bearing

(mm)

()

(m RL)

(m RL)

66.3

313.4

313.38

35 (2)

115 (2)

49.5

302.8

302.9

85 (2)

124 (1)

(m)

CE P

BH1A

Approx. Movement depth

D

Top of IncloTube

TE

Inclinometer

MA

Table 2. Summary of inclinometer results (July 2008 to October 2010, when the inclinometers were assumed to have sheared due to landslide movement). Inclinometer locations are shown on Fig. 3.

1

Depth of slip surface identified from logging of the drill holes. RL is the relative level from a given datum. 2 Equipment precision calculated from the equipment specifications (Slope Indicator 2005), and are based on measurements between two surveys, the reference survey 9/07/2008 and the most recent survey 2/10/2010. Uncertainties are estimated at 1 (68% confidence). 3 Estimates of the uncertainty are calculated using the least squares method at 1  (68% confidence).

AC

Table 3. Cumulative horizontal displacements calculated at the surface from CGPS stations UTK1, UTK2, UTK3 and UTK4 for the period 18/07/2008 to 4/02/2011 Number of days installed

Total cumulative horizontal displacement in period along 1 given bearing (mm)

Rate of displacement (mm/year)

Displacement 2 bearing ()

UTK1

928

268

105

155 ±(2)

UTK2

927

223

88

149 ±(2)

UTK3

931

182

71

142 ±(4)

UTK4

932

195

76

146 ±(1)

Location

1

2

Mean estimates of the uncertainty are about ±6 mm at 1 (68% confidence).

Estimates of the uncertainty are calculated using the least squares method at 1  (68% confidence).

Table 4. Summary of faster displacement periods recorded by the cGPS stations on the landslide (locations are shown in Fig. 3). Movement period

Time period

Cumulative horizontal displacements along main 1 movement bearings (mm) UTK1

UTK2

UTK3

UTK4

1

15/08/2008 to 22/08/2008

12

9

6

7

2

6/09/2008 to 6/11/2008

43

2.2

0.5

2.6

31

ACCEPTED MANUSCRIPT 30/09/2009 to 10/10/2009

7

11

7

6

4

16/09/2010 to 29/09/2010

110

168

146

138

Main movement bearings and measurement precision are shown on Fig. 6.

T

1

3

1

1

± (mm/day)

UTK1

23

0.2

UTK2

13

0.1

UTK3

3

0.1

UTK4

15

0.1

N daily records

77 to 3

0.2 to 4.0

815

8

0.6

878

5

0.5

879

5

0.3

879

1

Vertical rate (mm/day)

± (mm/day)

SC R

Horizontal rate (mm/day)

NU

cGPS station

IP

Table 5. Summary of slow displacement periods detrended by removing periods of faster displacement (cGPS station locations are shown on Fig. 3).

Estimates of the uncertainty are calculated using the least squares method at 1  (68% confidence)

UTK3

CGPS

BH3A

Inclinometer

1

Rate of displacement (mm/year)

Displacement 2 bearing ()

25 (6)

13

142 (7)

15 (2)

8

124 (1)

Total displacement (mm)

D

Equipment type

TE

Location

MA

Table 6. Comparison between the cumulative horizontal displacements calculated at the surface from CGPS station UTK3 and those at depth along the slip surface from inclinometer BH3A, for the period 9/07/2008 to 13/05/2010

Estimates of the uncertainty at 1 (68% confidence)

2

Estimates of the uncertainty are calculated using the least squares method at 1  (68% confidence).

AC

CE P

1

32

ACCEPTED MANUSCRIPT FIGURE CAPTIONS

Figure 1.

The location of the Utiku landslide in New Zealand, and the extent of exposed

T

Neogene-age sedimentary rocks. Aerial oblique view of the Utiku landslide taken in 1965 (source: L. Homer)

Figure 3.

The engineering geology of the Utiku landslide. The topographic contour

IP

Figure 2.

SC R

interval (thin grey lines) is 2 m and are derived from 2005 photogrammetry. Coordinates and bearings in terms of Geodetic Datum 2000. Long section A-A’ and B-B’ through the Utiku landslide (locations of section lines are shown on Fig. 3). Figure 5

NU

Figure 4.

Selected inclinometer cumulative displacement plots for BH1A and BH3A for

MA

their main axes (A0). Measurements were made at 0.5 m increments. Inclinometer A0 references have been corrected to grid north (Magnetic plus

D

21). The readings have been plotted versus depth (m) measured from the

TE

top of the inclinometer casing. The cumulative displacements are referenced from the bottom of the tube. The plotted point at any depth is the sum of

Figure 6

CE P

incremental lateral deviations up to and including that depth. The magnitude and bearing of horizontal surface displacements for the historical survey marks and the continuous GPS (cGPS) stations located on

AC

the Utiku landslide. The historical displacement data is for the period 1965 to 1972, and the cGPS data is for the period 2008 to 2011. Data shown are movement rates (velocities) and uncertainties (at 1 , 68% confidence) in m/yr (historical data) and mm/yr (cGPS data).

Figure 7

Historical surface movement of the Utiku landslide and associated rainfall. A) Cumulative horizontal displacement of selected historical survey marks (locations are shown in Fig. 6), plotted along their main displacement bearings. B) Daily rainfall and cumulative deviation from mean daily rainfall recorded at the Rangitikei District Council gauge at Taihape (about 6 km north of the Utiku landslide).

Figure 8

Continuous GPS daily positions, showing the “raw” and “filtered” deviations from mean horizontal East- and North-coordinate values in mm calculated for 33

ACCEPTED MANUSCRIPT each 24-hour epoch and their uncertainties (at 2, 95% confidence), for selected cGPS stations (locations are shown in Fig. 3). Data are colour-coded by time period. The “raw” mean horizontal East and North coordinates and

T

values (observations), calculated for each 24-hour epoch data were regionally

IP

“filtered” (following the methodology of Wallace and Beavan, 2010) to remove

SC R

common-mode noise and plate-tectonic displacements, by referencing their motions to cGPS station THAP (A) located 6 km to the north of the landslide. The local stability of THAP was verified by a 2nd reference station UTKU (B),

NU

0.5 km east of the landslide. C) “raw” values for UTK1, and D) “filtered” values for UTK1.

Daily “filtered” cumulative horizontal displacements of cGPS stations (UTK1

MA

Figure 9

to 4) located on the Utiku landslide (locations are shown in Fig. 3), plotted along their main displacement bearings. FD – fast displacement periods. SD

D

– slow displacement periods. Note: the cumulative displacement cGPS time

Daily movement data for selected cGPS stations during fast displacement

CE P

Figure 10

TE

all originate at zero and have been offset from zero for presentation purpose.

(FD) periods 2 and 4 (locations are shown in Fig. 3).

A) Horizontal

displacement for FD period 2; B) Horizontal displacement for FD period 4; C)

AC

Velocity for FD period 2) D: Velocity for FD period 4) E: Acceleration for FD period 2; F) Acceleration for FD period 4; Note that the time scales for each movement period vary. Smoothed values are derived using a Gaussian smoothing kernel.

Figure 11

Daily horizontal and vertical displacements of selected cGPS stations (locations are shown in Fig. 4) filtered by removing the displacements recorded during the periods of fast displacement. Linear trend-lines are fitted to the data and the mean rates of slow displacement (vertical and horizontal) and the uncertainties (at 1, 65% confidence) in mm/yr are given. The total displacements (at the top of each chart) are calculated from the gradients of the horizontal and vertical motion time series, and the angles represent the translational angles of displacement from horizontal.

34

ACCEPTED MANUSCRIPT Figure 12

Seasonally cyclic

slow

displacements.

Daily

horizontal

and

vertical

displacements of selected cGPS stations (locations are shown in Fig. 3) were filtered (detrended) by removing the displacements recorded during periods

Daily pore pressure trends in kPa for selected piezometers BH1, BH3 and

IP

Figure 13

T

of fast displacement (FD) and slow displacement (SD).

barometric changes in pressure. Figure 14

SC R

PZA (locations are shown in Fig. 3). The data are filtered to remove

The daily changes in pore pressure (A to C) and the deviations from mean

NU

pore pressure (D to F) for piezometers BH1, BH3 and PZA (locations are shown in Fig. 3).

Results from the pore pressure and antecedent rainfall analysis. The pore

MA

Figure 15

pressure at a given time and the accumulated antecedent rainfall were assessed incrementally for each piezometer, using the daily rainfall and pore

D

pressure values recorded on the landslide. The correlation analysis assumes

Figure 16

TE

a linear relationship between pore pressure and rainfall. Daily pore pressures recorded at piezometers PZA and BH3 and cumulative

CE P

horizontal displacements recorded at cGPS stations UTK1 and UTK3 (A and B) (locations are shown in Fig. 3). C) and D) show the daily pore pressures

AC

plotted against the corresponding daily horizontal displacement of the selected piezometers and cGPS stations. Route logs (chronological order of the data, oldest to youngest) are derived using a Gaussian smoothing kernel. FD – fast displacement periods.

Figure 17

Fast displacement period 4 (FD4). A) Displacement measured at UTK1 and pore pressures measured at piezometer PZA. Measurements represent daily readings. B) Displacement rate of UTK1 plotted against the corresponding pore pressure from piezometer PZA for FD period 4.

Figure 18

Daily pore

pressure

and corresponding detrended daily cumulative

displacements of adjacent cGPS stations showing seasonally cyclic slow displacement.

35

ACCEPTED MANUSCRIPT Figure 19

Conceptual model illustrating: A) displacement patterns; and B) relationship between pore pressure, movement patterns and mechanisms, of deep-seated

AC

CE P

TE

D

MA

NU

SC R

IP

T

translational rock slides as represented by the Utiku landslide.

36

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

37

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

38

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

39

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

40

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

41

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

42

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

43

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

44

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

45

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

46

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

47

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

48

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

49

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

50

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

TE

D

Fig 15

51

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

52

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

53

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

54

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

55

ACCEPTED MANUSCRIPT Highlights



Relationships between landslide motion and its causes have been studied



Displacements of the 2210 m Utiku landslide, New Zealand, were measured using

3

T

6

This landslide represents one of 7,000 mapped landslides of this type in

SC R



IP

continuous GPS

sedimentary rocks 

Displacement comprised patterns of episodic post failure creep punctuated by

A framework for the movement of reactivated landslides in similar settings has been

CE P

TE

D

MA

proposed

AC



NU

intervals of rest

56