GNS Science

The Kaikoura landslide story: it’s not over yet, but its not all bad news!

GNS Science

Contents

• Landslides triggered by the MW 7.8 14 November 2016 Kaikoura EQ

• The post-EQ

response of the slopes, – Dams; – Slopes; and – Rivers.

• How we use this information for the benefit of NZ

Ohau Point

1) 2)

1) 2)

GNS Science

The landslide inventory and published ground motion simulations

N landslides > 25,000 (as of August 2018)

PGV* Holden et al. 2017)

GNS Science

Hapuku rock avalanche

• Several fault ruptures in source – source area discontinuity controlled in greywacke

• Largest mapped landslide

• Dammed the Hapuku River

• Elevation of crown 2,800 m AMSL

• Vol. ero: 13 (±2) M m3

• Vol. dep: 20 (±2) < m3 • Runout about 2.7 km

GNS Science

Hapuku rock avalanche

• Difference model based on post EQ digital surface model (DSM) minus pre EQ DSM

GNS Science

Seafront Rock slide/slump

• Dominant material: Paleogene Limestone (bedding steeply dipping into slope)

• Failure surface: Through rock mass. Surface fault ruptures pass though source

• Setting: Coastal slopes (up to 430 m AMSL) • Volume range: 8–12 M m3

• Runout up to 1 km

GNS Science

The main controls on the landslide distribution

• Variables investigated (12) • Variables with statistical significance:

– Slope angle – Distance to surface fault rupture – Elevation – Geology – PGA or PGV – LSR: Local Slope Relief

It’s not just about the faults! Although they are important

Faults

Landslides

GNS Science

Distance to surface fault rupture has the strongest influence

• Seven of the largest eight landslides had surface fault ruptures pass through their source areas

• Most landslides occurred within a few kilometres of a fault that ruptured

Fault rupture

GNS Science

Kaikoura EQ landslide observations

• The number of landslides with source areas ≥10,000 m2 triggered by this EQ is smaller than those (of similar size) triggered by similar magnitude EQ’s in NZ

• Seven of the largest eight landslides (from 5 to 20 x 106 m3) occurred on faults that ruptured to the surface during the earthquake

• The landslide density within 200 m of a mapped surface fault rupture is three times that at a distance of 2,500 m or more from a mapped surface fault rupture

• The “distance to fault” predictor variable, when used as a proxy for ground-motion intensity, and when combined with slope angle, geology and elevation, has more predictive power than PGA or PGV

• Geology controls the types of landslide and failure mechanism

Massey et al. (2018) BSSA Kaikoura special edition

GNS Science

Post-EQ response of the slopes and rivers

GNS Science

Cyclones Cook and Debbie and the dams

• Washed away most of the landslide dams, including 10 of the top 11

• Remobilisation of debris from landslides

Map showing the locations of the 200 significant landslide dams

Hapuku dam

GNS Science

Assessment of landslide dam probability of failure

Conway 420Hapuku 740

Linton 340Ote Makura 100 Clarence

100

1000

10000

100000

1000000

10000000

100000000

1E+09

1 10 100 1000 10000

Dam

vo

lum

e [m

3 ]/

Max

dam

he

igh

t [m

]

Upstream catchment area [km2]

0% probability of failure

25% probability of failure

50% probability of failure

75% probability of failure

90% probability of failure

Hapuku before Debbie

Hapuku after Debbie

Probability of dam failure: regional-scale assessment based on dam geometry and catchment area (stream power) = The Dimensionless Blockage Index (DBI)

GNS Science

Example dam breach models and field verification – The Hapuku

• Repeat surveys of dam pre- and post-overtopping

• Geotechnical characterisation of the materials forming the dam

• Assessment of dam failure modes and debris flood modelling

• Modelled debris heights checked against actual post-breach debris heights

• Used to forecast flood and debris inundation zones downstream

GNS Science

Ex Tropical Cyclone Gita rain and impacts

• Max recorded rain along the Kaikoura coast for different time periods was: – 1 hour = 54 mm – 12 hour = 270 mm – 24 hour = 302 mm

• An estimated 300,000 m3 of debris covered the road and rail at multiple sites (source NCTIR)

• Road and rail closed for about 10 days • Several homes were hit by debris and the

road and rail were closed prior to the storm

GNS Science

0

100

200

300

400

500

600

0 10 20 30

Rai

n h

eig

ht

(mm

)

Rain duration (hours)

Kaikoura coast rainfall 20 February 2018

0

100

200

300

400

500

600

0 10 20 30

Ra

in h

eig

ht

(mm

)

Rain duration (hours)

Kaikoura coast rainfall Cyclone Alison 1975

The rain: Gita 2018 and Alison 1975

A) B)

GNS Science

So what was the return period of the rain?

1

10

100

1000

0 10 20 30

Re

turn

pe

rio

d (

ye

ars)

Rain duration (hours)

Kaikoura coast rainfall 20 February 2018

1

10

100

1000

0 10 20 30

Re

turn

pe

rio

d (

ye

ars)

Rain duration (hours)

Kaikoura coast rainfall Cyclone Alison 1975

A) B)

GNS Science

The landslides

• Most of the landslides were highly

mobile debris flows. • Most occurred in locations where

slopes had been damaged as a result of the MW7.8 14 November 2016 earthquake, comprising: – Remobilisation of debris from landslides

triggered by the EQ – Enlargement of the source areas of

landslides triggered by the EQ – Remobilisation of landslide debris in

streams

GNS Science

Debris flows: Reactivation of EQ-induced landslide debris

GNS Science

River response

-500000

-250000

0

250000

500000

0 2500 5000 7500 10000 12500 15000 17500 20000

Vo

lum

e T

ran

sfe

rre

d

Hapuku - Downstream Transfer

2016-2017 2017-2018

2016-2017

2017-2018

Net Erosion

Net Deposition

Hapuku River response

Red = erosion Blue = deposition

Difference map, 2016-2018 LIDAR surveys for the Kowhai River.

Difference Map, 2016-2018, Kowhai R.

Bank Erosion, Channel Migration

Gravel Extraction Works

Cumulative Downstream Net Total of Erosion/Deposition

Active Channel Bounds

Noise Threshold

• The annual sediment budget for the rivers varies quite a bit;

• upstream sediment inputs from e.g., landslides can keep the system in a surplus state (i.e., lots of debris available to erode and transport) for a few years

• This can involve much more dynamic channel behaviour, including channel switching and adjustment of river course.

• Gravel extraction, at 2017 rates, appears to be quite effective in managing the current surplus levels of sediment in the Kowhai.

• It will be interesting to monitor the balance between headwater supply and downstream extraction over the coming years

Some prelim observations: Rivers

GNS Science

Should ALL this be expected after a major EQ?

YES

Port Hills slope response 2010 to 2016

Redcliffs surface change models following the main Canterbury EQ’s, derived from repeat ground-based LIDAR surveys

Redcliffs

Ohau Point

Port Hills slope response 2010 to 2016

4/09/2010 EQ

22/02/2011 EQ

16/04/2011 EQ

13/06/2011 EQ

23/12/2011 EQ14/02/2016 EQ

0.01

0.1

1

10

100

1000

10000

0 500 1,000 1,500 2,000

Vo

lum

e f

alli

ng

fro

m t

he

cli

ff (m

3 /d

ay)

Days after 4/09/2010 Darfield Earthquake

EQ Rockfall rates from TLSand Airborne LiDARsurveys

Non EQ Rockfall ratesfrom TLS surveys

Mean estimate of the pre2010-2011 Canterburyearthquake rockfall rate

Upper estimate of the pre2010-2011 Canterburyearthquake rockfall rate

Power (Non EQ Rockfallrates from TLS surveys)

• Non-EQ rockfall rates reached the longer-term average rates after about 5-6 years. Porthills slopes are a lot smaller than those in Kaikoura

• In Wenchuan, China following the 2008 M 7.8 EQ, the slopes took about 6-10 years to recover. These slopes are more comparable to those in Kaikoura

• But it depends on what happens now??

Landslide losses in NZ: What are we going to do about it?

• Landslides cost New Zealand an average of $300M/yr (Source: Page, 2013)

• Individual landslide hazard events from rainstorms range from large events costing ~$350M (e.g. Cyclone Bola) to small events of ~$3.5M (Source: Page, 2013)

• Kaikoura losses $500M (Source MBIE) mainly landslide related

• For earthquake-induced landslides, the residential losses in the Port Hills were estimated to be $330M (Source: CERA). In Wellington they are expected to be $1B or more (Source EQC)

• They have killed more than 800 people since about 1880 (avg 5/yr), since 1910 about 3/yr as people have “naturally” migrated from being in high risk environments e.g., mining (Source GNS

Science)

• They are the frequent menace!

The efficacy of avoiding the problem

Post 22 Feb 2011

Post 14 Feb 2016

Whitewash Head, Port Hills, Christchurch

The efficacy of avoiding the problem

• Ohau point – road realigned away from the toe of the cliff • But it’s no always possible to avoid, in which case we have to either

mitigate or accept the risk

Photo: NZTA

Using this information for the benefit of NZ

GNS Science

The Problems:14 Nov 2016 MW 7.8 Kaikoura EQ and landslides

• Ground motion models evolved with time and could not be used to initially focus aerial surveys

• And as a result, the initial landslide forecast models were not useful

• Later models were also not that useful!

Allstadt et al., 2018 (BSSA)

GNS Science

Prototype Earthquake-induced Landslide (EIL) forecast tool for NZ

The solution: develop a landslide forecast tool that can provide advisory information soon after an earthquake to help focus reconnaissance and response efforts

GNS Science

a) Instrument PGA (streamed data sampled 5 mins after EQ), bounding polygon – PGA 20%g, 8km

grid b) Elevation (mAMSL), 32 m grid c) Slope angle (deg), 32 m grid

Landslide model variables: Example using the 2016 Kaikoura EQ

GNS Science

Landslide model variables: Using the 2016 Kaikoura EQ

d) Local slope relief (height), used as an amplification factor, 32 m grid e) Proximity to active fault, 32 m grid f) Geology types, 32 m grid

GNS Science

Advisory: Text and Maps 5-7 mins after tool triggered

TA that might be affected by landslides

Area where PGA exceeds 0.2 g (km2)

Percent of TA's area where PGA exceeds

0.2 g

Maximum PGA within TA (g)

Hurunui District 4205.08 48.6 0.68

Kaikoura District 2046.78 100 1.27

Lower Hutt City 1.2 0.3 0.22

Marlborough District 6362.82 60.8 1.27

Wellington City 22.72 7.8 0.26

Landslide probability model used is an updated version of the model presented in Massey et al., 2018 (BSSA)

GNS Science

EQ Landslide probability model for Wellington (adopting median PGA for Wellington Fault EQ)

Landslide runout model for Wellington

(adopting debris runout F-angles of 31º

(avalanche) and 21º (flows), projected from landslide source areas

where P≥50%)

Better land use planning: Future applications of the tool: including landslide runout models

• Apply hybrid landslide forecast model to Wellington materials and slope morphologies to identify potential source areas

• Apply runout models from source areas

Slippage Falling

debris

F-angle

GNS Science

Thanks to everyone who has contributed

This work was funded by GeoNet – New Zealand’s geological monitoring agency, the Natural Hazards Research Platform and the MBIE Endeavour programme The team: • GNS Science: Chris Massey, Sally Dellow, Simon Cox, Garth

Archibald, John Begg, Zane Bruce, Jon Carey, Fernando Della Pasqua, Phil Glassey, Katie Jones, Biljana Lukovic, Barbara Lyndsell, Brenda Rosser, Corinne Singeisen, Dougal Townsend, Pilar Villamor, Yoshi Koneko; Caroline Holden; Anna Kaiser; David Rhoades

• Massey University: Sam McColl • Auckland Uni: Jon Tunnicliffe • Canterbury University: Jonathan Davidson, Marlene Villeneuve,

Brendon Bradley • Victoria Uni: Jamie Howarth • Sheffield University, UK: David Petley • USGS landslide team: Jonathan Godt, Randall Jibson, Kate

Allstadt, Francis Rengers • GEER landslide team: Joseph Wartman, Ellen Rathje, Nick Sitar,

Athanasopoulos-Zekkos, Adda, John Manousakis, Michael Little.