Predicting Riverbank Collapse Using Numerical Modelling

Mark Jaksa, Yien Lik Kuo, Chen Liang, Bertram Ostendorf, Tom Hubble and Elyssa De Carli

Goyder Institute for Water Research Technical Report Series No. 15/41

www.goyderinstitute.org Riverbank Collapse Task 5 Report

Goyder Institute for Water Research Technical Report Series ISSN: 1839-2725

The Goyder Institute for Water Research is a partnership between the South Australian Government through the Department for Environment, Water and Natural Resources, CSIRO, Flinders University, the University of Adelaide, the University of South Australia and ICE WaRM (the International Centre of Excellence in Water Resources Management). The Institute will enhance the South Australian Government’s capacity to develop and deliver science-based policy solutions in water management. It brings together the best scientists and researchers across Australia to provide expert and independent scientific advice to inform good government water policy and identify future threats and opportunities to water security.

The following organisation contributed to this report:

Enquires should be addressed to: Goyder Institute for Water Research Level 4, 33 King William St, Adelaide SA 5000 tel: (08) 8236 5209 e-mail: [email protected]

Citation Jaksa, MB, Kuo, YL, Liang, C, Ostendorf, B, Hubble, T, and de Carli, E. 2015, Predicting Riverbank Collapse Using Numerical Modelling, Goyder Institute for Water Research Technical Report Series No. 15/41, Adelaide, South Australia

Copyright © 2015 University of Adelaide. To the extent permitted by law, all rights are reserved and no part of this publication covered by copyright may be reproduced or copied in any form or by any means except with the written permission of University of Adelaide.

Disclaimer The Participants advise that the information contained in this publication comprises general statements based on scientific research and does not warrant or represent the completeness of any information or material in this publication.

Page 2 of 77 Riverbank Collapse Task 5 Report

Table of Contents Executive Summary...... 8

1 Introduction ...... 10

2 GIS-based Analysis of Riverbank Collapses at Long Island Marina and Tailem Bend ...... 12 2.1 Introduction ...... 12 2.2 Background and Methodology ...... 14 2.3 Results ...... 18 2.4 Summary ...... 19

3 Back Analysis of Historical Riverbank Collapses ...... 20 3.1 Introduction ...... 20 3.2 Methodology ...... 22 3.3 Results ...... 29 3.4 Summary ...... 33

4 Effects of River Level Fluctuation and Climate on Riverbank Stability ...... 34 4.1 Background ...... 34 4.2 Study Area ...... 36 4.3 Methodology ...... 38 4.4 Results ...... 47 4.5 Summary ...... 55

5 Identifying the Areas Susceptible to Riverbank Instability ...... 56 5.1 Introduction ...... 56 5.2 Methodology and Results ...... 56 5.3 Summary ...... 65

6 Summary ...... 68

7 References ...... 69

Page 3 of 77 Riverbank Collapse Task 5 Report

List of Figures

FIGURE 1: LOCATION OF THE STUDY AREA...... 12 FIGURE 2: LONG ISLAND MARINA STUDY SITE: (A) LOCATIONS OF 5 SIGNIFICANT FAILURES; (B) LOCATION PLAN OF IN SITU TESTING AND RECORDED COLLAPSES; AND (C) DISTRIBUTION OF BANK CROSS-SECTIONS IN 3D PERSPECTIVE...... 13 FIGURE 3: EXAMPLES OF VISUAL INTERPRETATION OF 2008 AND 2010 AERIAL IMAGES USING ARCGIS AT: (A) MURRAY BRIDGE; AND (B) TAILEM BEND...... 13 FIGURE 4: RIVER MURRAY WATER LEVEL AT MURRAY BRIDGE 1/12/1986 TO 11/07/2011 (BASED ON DFW, 2010)...... 15 FIGURE 5: SCHEMATIC OF THE LOCATIONS AND THE SELF-WEIGHT OF THE STRUCTURES ON THE RIVERBANK...... 16 FIGURE 6: MINIMUM FOS AND POTENTIAL SLIP SURFACE OF A DEEP-SEATED ROTATIONAL FAILURE OBTAINED FROM SVSLOPE AT LOCATION NO. 21 WHEN THE WATER LEVEL WAS AT 0 M AHD. .. 16 FIGURE 7: BACK ANALYSES USING THREE GEOTECHNICAL MODELS...... 17 FIGURE 8: FACTORS OF SAFETY OF NEIGHBOURING CROSS SECTIONS (0 AND 0.5 M AHD)...... 18 FIGURE 9: PREDICTIONS OF RIVERBANK SUSCEPTIBILITY WITH RIVER LEVELS AT (A) 0 M AHD AND (B) 0.5 M AHD...... 19 FIGURE 10: THE GEOGRAPHICAL LOCATIONS OF THE FOUR STUDIED SITES AND MAJOR RIVERBANK COLLAPSES IN THE PAST, AS WELL AS THE PROXIMITIES OF THE BOREHOLE AND RIVER LEVEL OBSERVATION STATIONS...... 21 FIGURE 11: ADOPTED VISUAL INTERPRETATION METHOD OF HIGH-RESOLUTION AERIAL IMAGES: (A), (C), (E) AND (G) ARE AERIAL PHOTOGRAPHS ACQUIRED IN MARCH 2008 AT MN, WR, RFR AND WS, RESPECTIVELY; (B), (D), (F), AND (H) ARE AERIAL PHOTOGRAPHS ACQUIRED IN MAY 2010 AT MN, WR, RFR AND WS, RESPECTIVELY...... 23 FIGURE 12: EXAMPLE OF ADOPTED ELEVATION COMPARISON METHOD ON DEMS AT WR (A) 1 M RESOLUTION DEM ACQUIRED IN 2008; (B) 0.2 M RESOLUTION DEM ACQUIRED IN 2010)...... 24 FIGURE 13: RESULTS OF CPTU SOUNDING (A) PROFILE; AND (B) PORE PRESSURE DISSIPATION TEST RESULTS...... 27 FIGURE 14: DAILY RIVER LEVELS AND DAILY RAINFALL RECORDED AT (A) MANNUM (MN) SITE IN APRIL 2009; (B) WOODLANE RESERVE (WR) SITE IN FEBRUARY 2009; (C) RIVER FRONT ROAD, MURRAY BRIDGE (RFR) SITE IN FEBRUARY 2009; AND (D) WHITE SANDS (WS) SITE IN APRIL 2009...... 28 FIGURE 15: RIVERBANK STABILITY ANALYSIS OF THE MANNUM (MN) SITE ON 21 APRIL 2009...... 30 FIGURE 16: RIVERBANK STABILITY ANALYSIS OF THE WOODLANE RESERVE (WR) SITE ON 26 FEBRUARY 2009...... 30 FIGURE 17: RIVERBANK STABILITY ANALYSIS OF THE RIVER FRONT ROAD, MURRAY BRIDGE (RFR) SITE ON 6 FEBRUARY 2009...... 30

Page 4 of 77 Riverbank Collapse Task 5 Report

FIGURE 18: RIVERBANK STABILITY ANALYSIS OF THE WHITE SANDS (WS) SITE ON 23 APRIL 2009...... 31 FIGURE 19: RIVERBANK COLLAPSE FACTOR OF SAFETY TIME SERIES FOR MANNUM (MN) IN APRIL 2009...... 31 FIGURE 20: RIVERBANK COLLAPSE FACTOR OF SAFETY TIME SERIES FOR WOODLANE RESERVE (WR) IN FEBRUARY 2009...... 32 FIGURE 21: RIVERBANK COLLAPSE FACTOR OF SAFETY TIME SERIES FOR RIVER FRONT ROAD (RFR) IN FEBRUARY 2009...... 32 FIGURE 22: RIVERBANK COLLAPSE FACTOR OF SAFETY TIME SERIES FOR WHITE SANDS (WS) IN APRIL 2009...... 32 FIGURE 23: DETAILS OF THE LONG ISLAND MARINA SITE...... 37 FIGURE 24: RIVERBANK GEOMETRY DEFINITION...... 40 FIGURE 25: EXAMPLE OF ADOPTED VISUAL INTERPRETATION PROCESS ON HIGH-RESOLUTION, AERIAL IMAGES WITHIN ARCGIS...... 41 FIGURE 26: GEOTECHNICAL PROFILES BASED ON SOIL SAMPLES TAKEN FROM SR-BH1 AND SR-CPTU6S AT LONG ISLAND MARINA...... 42 FIGURE 27: PARTICLE SIZE DISTRIBUTIONS BASED ON THE SOIL SAMPLES FROM FOUR DIFFERENT DEPTHS IN BOREHOLE SR-BH1...... 42 FIGURE 28: ESTIMATED SWCCS FOR THE THREE SOIL LAYERS AT LONG ISLAND MARINA USING THE FREDLUND AND XING (1994) ESTIMATION METHOD...... 43 FIGURE 29: DAILY RIVER LEVELS, DAILY RAINFALL AND DAILY MEAN TEMPERATURE FROM MAY 1, 2008 TO FEBRUARY 28, 2009 AT LONG ISLAND MARINA...... 45 FIGURE 30: RESULTS OF (A) 2D (DAY 282: FEBRUARY 6, 2008) AND (B) 3D (DAY 287: FEBRUARY 10, 2008) RIVERBANK STABILITY ANALYSES OF LONG ISLAND MARINA SITE...... 46 FIGURE 31: EVOLUTION OF PORE WATER PRESSURE AT 5 SELECTED NODES THROUGH THE ENTIRE RESEARCH PERIOD ACCOUNTING FOR, AND WITHOUT, EVAPORATION...... 48 FIGURE 32: PWP DISTRIBUTIONS AS A RESULT OF (A) THE HIGHEST (DAY 138) AND (B) LOWEST (DAY 302) RIVER LEVELS...... 49 FIGURE 33: FACTORS OF SAFETY FROM THE 2D, 3D AND CRLM MODELS...... 50 FIGURE 34: FACTORS OF SAFETY FOR HISTORICAL MODEL (HM) AND CONSTANT RIVER STAGE MODEL (CRLM) IN TWO SCENARIOS...... 53 FIGURE 35: MAGNIFIED RAINFALL MODEL (MRM) UNDER DIFFERENT RIVER LEVEL SCENARIOS...... 54 FIGURE 36: LOCATING DEEP HOLES NEAR THIELE RESERVE USING: (A) BATHYMETRY FROM HUBBLE AND DE CARLI (2015) (AREAS OF BEDROCK MARGINS ARE REPRESENTED BY DASHED WHITE LINE); (B) A CONTOUR MAP OBTAINED FROM ARCGIS SUPERIMPOSED ON SATELLITE IMAGERY FROM GOOGLE MAPS; (C) BLUE CONTOUR LINES SHOWING ELEVATIONS THAT ARE –10 M AHD OR DEEPER, WHILST THE GREEN CONTOUR LINES REPRESENT 0 M AHD...... 57

Page 5 of 77 Riverbank Collapse Task 5 Report

FIGURE 37: LOCATIONS OF DEEPENED CHANNEL BEDS DOWNSTREAM FROM BLANCHETOWN TO WALKER FLAT (BLUE CONTOUR LINES SHOW ELEVATIONS THAT ARE AT –10 M AHD OR DEEPER, WHILST THE GREEN CONTOUR LINES ARE AT 0 M AHD)...... 58 FIGURE 38: LOCATIONS OF DEEPENED CHANNEL BEDS DOWNSTREAM FROM WALKER FLAT TO MURRAY BRIDGE EAST (BLUE CONTOUR LINES SHOW ELEVATIONS THAT ARE AT –10 M AHD OR DEEPER, WHILST THE GREEN CONTOUR LINES ARE AT 0 M AHD)...... 59 FIGURE 39: LOCATIONS OF DEEPENED CHANNEL BEDS DOWNSTREAM FROM MURRAY BRIDGE EAST TO WELLINGTON (BLUE CONTOUR LINES SHOW ELEVATIONS THAT ARE AT –10 M AHD OR DEEPER, WHILST THE GREEN CONTOUR LINES ARE AT 0 M AHD)...... 60 FIGURE 40: A SERIES OF TRANSECTS ALONG 0 M AHD CONTOUR LINES NEAR THIELE RESERVE (BLUE CONTOUR LINES SHOW ELEVATIONS THAT ARE AT –10 M AHD OR DEEPER, WHILST THE GREEN CONTOUR LINES ARE AT 0 M AHD. THE RED DOT MARKS THE APPROXIMATE LOCATION OF THE PAST RIVERBANK INSTABILITY INCIDENT AND THE DASHED ORANGE LINE MARKS THE AREAS OF THE BEDROCK MARGINS)...... 61 FIGURE 41: AN EXAMPLE OF AN SVSLOPE2D ANALYSIS (TRANSECT NO. 4212)...... 63 FIGURE 42: AN EXAMPLE OF THE RESULT OF AN SVSLOPE2D ANALYSIS (TRANSECT NO. 5529)...... 63 FIGURE 43: RIVERBANKS THAT ARE VULNERABLE TO SLUMPING ARE IDENTIFIED (SHOWN IN YELLOW DOTS) AFTER THE ANALYSES (ASSUMING A RIVER LEVEL AT 0 M AHD; BLUE CONTOUR LINES SHOW ELEVATIONS THAT ARE AT –10 M AHD OR DEEPER WITH ONE METRE INTERVAL; RED DOT SHOWS THE APPROXIMATE LOCATION OF THE OCCURRENCE OF PAST SLUMPING; DASHED YELLOW LINES ARE AREAS OF BEDROCK MARGINS; AND LEVEES ARE REPRESENTED BY THE THICK GREEN LINE)...... 64 FIGURE 44: INCLINATION OF THE RIVERBANKS NEAR THIELE RESERVE USING A 1 × 1 MATRIX OF GRIDS...... 66 FIGURE 45: INCLINATION OF THE RIVERBANKS NEAR THIELE RESERVE USING A 10 × 10 MATRIX OF GRIDS...... 67

Page 6 of 77 Riverbank Collapse Task 5 Report

List of Tables

TABLE 1: NATURE OF THE HISTORICAL RIVERBANK COLLAPSE-RELATED INCIDENTS AT 4 STUDY SITES...... 20 TABLE 2: SOIL PROPERTIES FOR SATURATED AND UNSATURATED FLOW MODELLING...... 26 TABLE 3: GEOTECHNICAL MODELS OF THE CLAY LAYER OBTAINED FROM BACK ANALYSES...... 29 TABLE 4: MODEL VALIDATION...... 31 TABLE 5: SOIL PARAMETERS FOR STABILITY ASSESSMENT...... 41 TABLE 6: EQUATIONS USED TO CALCULATE FREDLUND AND XING (1994) SWCC FITTING PARAMETERS BASED ON THE SOIL GRAIN SIZE DISTRIBUTION...... 44 TABLE 7: ADOPTED GEOTECHNICAL PARAMETERS FOR STABILITY CALCULATIONS...... 62

Page 7 of 77 Riverbank Collapse Task 5 Report

Executive Summary

Unprecedented low river levels in the Lower River Murray (downstream of Lock 1, Blanchetown, South Australia) during the Millennium Drought, which occurred between 2005 and 2010, adversely contributed to more than 162 riverbank collapse-related incidents and a long-term metastable condition to the riverbanks along the Lower River Murray, which have recently been considered as the dominating factors inducing bank collapse. A number of studies have recently been undertaken by the authors to examine the causes of these historical collapses, as well as developing a framework to identify regions along the Lower River Murray that have a potentially higher risk of riverbank collapse. This report summarises these riverbank collapse studies.

Back analyses of some of the larger past failures were undertaken at a number of sites, namely Long Island Marina near Murray Bridge; East Front Road between Mannum and Younghusband; Whitesands near Tailem Bend; and Woodlane Reserve. Based on government inventories, the collapsed riverbank sections were identified and vectorised using visual interpretation within the ArcGIS geographic information system (GIS) framework. With the aid of high-resolution aerial photographs, digital elevation models (DEMs) and ArcGIS, the two-dimensional (2D) models of past collapsed riverbank were established and analysed. A three-dimensional model (at Long Island Marina) was also analysed. The integration of GIS with high-resolution spatial data facilitates the process of the identification of collapsed regions, model geometry development and the calculation of the dimensions of the collapsed regions.

From the GIS and DEM data, the topographies of the riverbanks were input into the SVSlope slope stability analysis program, which was in turn linked to the SVFlux groundwater seepage modelling software in order to perform transient back-analysis of riverbank stability. The operating shear strengths of the soils were determined using back-analyses and these were found to be consistent with those obtained from laboratory and field testing, and reported previously. The studies have also demonstrated that the riverbank collapses can be accurately and reliably predicted by using the proposed frameworks. Both the adopted process of 2D and 3D stability modelling yielded excellent predictions of the collapses when compared against the recorded dates and dimensions of the failed regions.

In summary, the studies have shown that the unprecedented low river levels are the main cause of the past riverbank instability incidents. Transient analyses have shown that river fluctuations, rather than climatic factors, dominate the likelihood of riverbank collapse along the Lower River Murray. Sudden or rapid drawdown events (ratios of between –0.006 to –0.5 m per day) have been simulated and it has been observed that such drawdown events may lead to riverbank instability. Furthermore, extreme rainfall events (e.g. rainfall exceeding 120 mm per day, whilst unlikely but plausible) coinciding with river levels at between –0.2 m to –0.5 m AHD, are also likely to trigger riverbank

Page 8 of 77 Riverbank Collapse Task 5 Report collapses. Finally, this report also presents a framework for the preliminary identification of regions along the Lower River Murray which potentially demonstrate high susceptibility to riverbank collapse.

Page 9 of 77 Riverbank Collapse Task 5 Report

1 Introduction

In general, bank retreat is the combined result of various influencing factors, such as bank geometry (i.e. topography, slip surface and material boundary locations), riverbank materials, external loads acting on the bank (i.e. vegetation and infrastructure) and pore water pressure (Filz et al., 1992). Back-analysis is an analytical methodology to assess the status of inherent influencing factors on a failed or failing slope (Filz et al., 1992). For slope stability analyses, at the time of failure, the details of influencing factors and the ways they interacted are often ambiguous (Abramson et al., 2002). Based on the assumption that the factor of safety (FoS) was equal to unity at the time of failure, a back-analysis calculation can be undertaken and a series of relevant parameters obtained for a specific location (Abramson et al., 2002).

Geographic information system (GIS) technology has greatly facilitated landslide research by providing various functions for spatial data management. The GIS-based technique is comprised of data capturing, handling, processing, analysing, integrating, simulating, visualising and modelling (Carrara et al., 1991a, 1995b; Guzetti et al., 1999; Wang et al., 2005). GIS-based geotechnical models simplify the process of quantifying the spatially distributed parameters influencing collapse (Xie et al., 2004). By processing and integrating high-resolution, remotely-sensed data (i.e. satellite imagery, aerial photos and digital elevation models [DEMs]), GIS-based geotechnical models have been used to assess bank stability at relatively large scales (Hong et al., 2007). As a result, the GIS method has been adopted throughout the study.

The stability of the riverbank is determined by means of the SVSlope (SoilVision, 2009b) slope stability analysis program and expressed in terms of the Factor of Safety (FoS). The FoS is determined by dividing the forces resisting movement by the forces driving movement. By this definition, a geotechnical structure, or in the present study a riverbank, with a FoS of exactly 1 will support only the design load and no more and will effectively be at the limit of stability. Any additional load, or reduction in strength, will cause the riverbanks to collapse. Section 2 of this report outlines the integration of GIS and SVSlope to analyse riverbank stability at Long Island Marina and Tailem Bend.

To understand the underlying physical mechanics of the riverbank collapses along the lower reaches of the River Murray, as well as, to quantify the effects of river level fluctuation and climatic influences, such as rainfall, wind, mean temperature and evaporation on riverbank stability, back-analysis was undertaken to model 4 historical and representative riverbank collapses. Four very high-risk sites identified by Sinclair Knight Merz (SKM) (2010a) namely, East Front Road near Mannum; Woodlane Reserve; River Front Road near Murray Bridge; and White Sands, were selected for these analyses. The back-analyses were undertaken by combining the GIS method, slope stability analysis using limit equilibrium (SVSlope) and finite element simulation of groundwater seepage flow (SVFlux [SoilVision, 2009a]). The methodologies and the results of the transient analyses are presented in detail in Section

Page 10 of 77 Riverbank Collapse Task 5 Report

3 of this report. The results are shown as transient FoSs, that is fluctuation of FoS with historical river level variations and climatic data.

Section 0 presents the modelling of the riverbank collapse incident, in both 2D and 3D, which occurred at Long Island Marina, Murray Bridge, South Australia, on 4 Feb. 2009. The influence and sensitivity of river level fluctuations and climatic factors on riverbank stability are examined in detail. From the results of these analyses the dominant triggers of riverbank collapse were identified. Both 2D and 3D stability modelling yielded excellent predictions of the collapse when compared against the recorded date and dimensions of the collapsed region. The operating geotechnical parameters were obtained and are consistent with those derived from field investigations reported by Jaksa et al. (2015). It showed river fluctuations, rather than climatic factors, are the triggers of riverbank collapses along the Lower River Murray. Events such sudden or rapid drawdown and extreme rainfall were also simulated.

Further to the Task 4 report by Hubble and De Carli (2015), Section 0 identifies locations of deep holes in the channels of the Lower River Murray from Blanchetown to Wellington by analysing DEMs using the geoprocessing framework in ArcMap. Hubble and De Carli (2015) suggested that deep holes are formed due to either: (a) bedrock constriction and pronounced narrowing of the channel cross- section; or (b) large outcrops of bedrock which protrude up from the floor of the channel. Such geological features have generated erosive flow patterns during periods of higher flow that have eroded or scoured deep holes in the channel and over-steepened the channel margins and banks and past large failures were often associated with the these features. Deep holes are more common downstream of Mannum, where most of the large failures have occurred. Riverbanks near Thiele Reserve have been extensively analysed and the results showed that riverbanks with slope angles steeper than 40° are vulnerable and could collapse in the longer term, but the failure types are most likely to be shallow or planar, implying that, if they occur, relatively small volumes of soil will collapse into the river. Finally, the study showed that the slope inclination map is a very useful indicator for identifying riverbank instability near Thiele Reserve and will thus be explored further in future work.

Page 11 of 77 Riverbank Collapse Task 5 Report

2 GIS-based Analysis of Riverbank Collapses at Long Island Marina and Tailem Bend

2.1 Introduction This section demonstrates the use of back-analysis to assess the susceptibility of riverbank collapses of several failed sites at different river levels at Long Island Marina, Murray Bridge, South Australia. It presents the first effort in this research program to back-analyse the Long Island Marina failure, and was published previously by Liang et al. (2012). Section 3 builds upon these analyses and examines other sites where riverbank failures have occurred between 2005 and 2010 and Section 4 further extends and refines the Long Island Marina analyses and explores various river level and climatic scenarios.

The study area examined in the present section is located along the Lower River Murray near the Long Island Marina, Murray Bridge, South Australia, as shown in Figure 1. The historical inventory of riverbank collapse incidents (WaterConnect 2015), which is maintained by the South Australian Department of Environment, Water and Natural Resources (DEWNR) (formerly DFW), documents 9 riverbank collapse-related incidents occurred between 2008 and 2011. Of these, 4 were relatively minor (i.e. bank cracking, tree leaning and collapse), however, the remaining 5 were significant and are located adjacent to Long Island Marina, as shown in Figure 2(a).

Figure 1: Location of the study area.

For the purpose of back analysis, the collapsed sections of riverbank need to be identified with relatively high accuracy. With reference to a series of imprecise geographical coordinates of the collapsed bank sections from the DEWNR inventory, under the ArcGIS framework, a visual interpretation method was adapted using high-resolution aerial images taken at different periods to identify the collapse region. Based on the visual comparison of the aerial images taken in 2008, with a 5 m resolution, and in 2010, with a 0.5m resolution, the collapsed region was vectorised, as shown in Figure 3 by the dotted hatching.

Page 12 of 77 Riverbank Collapse Task 5 Report

There are shortcomings, however, when attempting to identify collapsed regions by means of visual interpretation. Specifically, it is extremely difficult to verify by eye whether a section of riverbank collapsed or is submerged beneath a higher river level. In order to resolve this issue, a digital elevation model (DEM) acquired in 2008 was used to validate the elevation of the section prior to collapse. By comparing the elevations from the DEM with the river levels at the date that the aerial photograph was taken (river level data were obtained from the observation point at Murray Bridge), collapsed regions were more accurately identified. Furthermore, the loss of large riparian plants and the construction of new infrastructure (e.g. jetty), observed from later aerial images, are also helpful in identifying the collapsed regions.

Figure 2: Long Island Marina study site: (a) locations of 5 significant failures; (b) location plan of in situ testing and recorded collapses; and (c) distribution of bank cross-sections in 3D perspective.

Figure 3: Examples of visual interpretation of 2008 and 2010 aerial images using ArcGIS at: (a) Murray Bridge; and (b) Tailem Bend.

Page 13 of 77 Riverbank Collapse Task 5 Report

2.2 Background and Methodology Riverbanks similar in material, groundwater and stratigraphy may behave significantly differently depending on their particular topographies (Abramson et al., 2002). Throughout this study, the geometries of the riverbanks were extracted from two DEMs (0.5 m and 5 m resolution, between the years 2007 and 2010), which were provided by DEWNR (WaterConnect, 2015), within the ArcGIS framework. As shown in Figure 2(b), 26 2D riverbank cross-sectional models were established and represented by 26 polyline slices. Each polyline consisted of 401 points in a line with a 10 m interval between neighbouring points. In order to conduct the back analyses, 5 cross-sectional models were selected to coincide with collapsed regions, which were vectorised previously. Based on the bilinear interpretation method, the elevation values were extracted from the DEMs and assigned to 401 points with singular points avoided.

Stratigraphic and geotechnical data for this research were collected from a series of site investigations (boreholes, and cone penetration, dilatometer, vane shear, standard penetration, pocket penetrometer and laboratory tests), performed by Sinclair Knight Merz (SKM, 2010a). Generally speaking, the soil profile comprises a 0.8 – 1.2 m thickness of silty sand (SM/SC) overlaying a silty clay, (unit weight of 18 ± 1kN/m3, internal angle of friction of 28 ± 2°, cohesion of 2 ± 2 kPa); 11 – 20 m thickness of dark grey, very soft, silty clay (CH) (unit weight of 16 ± 1 kN/m3, cohesion of 10 ± 5 kPa [10 kPa at the top, maximum of 25 kPa increasing by 1.25 kPa/m]); and a medium dense clayey sand (Monoman Sand) (SC/CL) layer (unit weight of 17 ± 1 kN/m3, internal angle of friction of 30 ± 2° and 2 ± 2 kPa cohesion).

Pore water pressure affects the riverbank stability by altering the shear strength of bank material (reduces the effective stress) and the self-weight of the soil mass (Sharma et al., 2002). It is well understood that positive pore water pressure, which plays a significant role in drawdown failures, reduces the effective shear strength of soils (Budhu and Gobin, 1995). In contrast, negative pore pressure, or soil suction, offers apparent cohesion, which stabilises the bank and manifests steeper banks. Casagli et al. (1997) stated that when the bank materials become saturated, collapses are more likely to occur. However, due to seasonal variations, negative pore water pressure do not contribute to long term stability (Sterrett and Edil 1982).

Fluctuations in river level directly affect the flow of water in and out of the bank and the pore water pressures within the bank (Green, 1999). Generally, over the last 2 decades or so, the river levels at Murray Bridge have fluctuated between 1.5 – 1.7 m AHD (Australian Height Datum), as shown in Figure 4. In general, the river level has remained relatively constant until the end of 2006, where it fell dramatically until October 2009, after which it has returned to its pre-2006 levels.

Page 14 of 77 Riverbank Collapse Task 5 Report

2

1.5

1

0.5

0

-0.5

Water Level (m AHD) (m Level Water -1

-1.5

-2

01-Dec-86 01-Dec-87 01-Dec-88 01-Dec-89 01-Dec-90 01-Dec-91 01-Dec-92 01-Dec-93 01-Dec-94 01-Dec-95 01-Dec-96 01-Dec-97 01-Dec-98 01-Dec-99 01-Dec-00 01-Dec-01 01-Dec-02 01-Dec-03 01-Dec-04 01-Dec-05 01-Dec-06 01-Dec-07 01-Dec-08 01-Dec-09 01-Dec-10 Date Figure 4: River Murray water level at Murray Bridge 1/12/1986 to 11/07/2011 (based on DFW, 2010).

Applied external loads are also known to affect slope stability. The external loads acting on riverbanks consist of two main categories: riparian plants and manmade structures. Considering the potential slip modes and the horizontal and lateral extent of the collapse, the impacts of failure were mainly determined by the dead weights and the distributions of the external loads (distances to the bank line).

The effects of riparian plants on riverbank stability are well known (e.g. Hickin, 1984, Thorne, 1990). The vegetation alters bank hydrology, flow hydraulics and the bank’s geotechnical properties (Abernethy and Rutherfurd, 2000). However, the effects of plants change with the seasons and their life cycle and, hence, are difficult to predict and integrate into bank stability analysis (Thorne and Osman, 1988; Abernethy and Rutherfurd, 2000). The self-weight of large vegetation, such as trees, and the effects of wind on the trees’ canopies, add load to the riverbank, which in turn, destabilises the bank. Combining the beneficial hydrological and mechanical stabilising effects from root suction and root matrix reinforcement, with the detrimental, additional load provided to the bank results in vegetation providing only a marginal influence on bank stability (Abernethy and Rutherfurd, 2000).

The location and the self-weight of manmade structures affect riverbank stability by imposing a surcharge on the bank. In the study area, a site inspection indicates that approximately 33% of houses are located quite close (less than 10 m) to the riverbank line. The load exerted by a house to the bank is roughly 20 kPa, which consists of the weight of the roof, walls and floor, as well as the live loads, as shown in Figure 5.

Using the data discussed above, a series of back-analyses were conducted using the limit equilibrium method within SVSlope (SoilVision, 2009b) assuming that a FoS of unity represents incipient failure.

Page 15 of 77 Riverbank Collapse Task 5 Report

Due to the distances between the site investigations and the actual locations of the collapse sites, the soil parameters collected from borehole logs were only applied as initial values in the FoS calculation.

Figure 5: Schematic of the locations and the self-weight of the structures on the riverbank.

The results of site investigations suggest that deep-seated rotational slips due to bank toe erosion in low flow situations, as well as slab failures, are the two dominant riverbank retreat modes along the Lower River Murray (Tajeddin et al., 2010; Hubble and De Carli, 2015). The deep-seated rotational failures typically occurred in the silty clay layer (between –5 to –20 m AHD), an example of which is shown in Figure 6.

Figure 6: Minimum FoS and potential slip surface of a deep-seated rotational failure obtained from SVSlope at Location No. 21 when the water level was at 0 m AHD.

Page 16 of 77 Riverbank Collapse Task 5 Report

The results of the back analyses are summarised in Figure 7: two geotechnical models were adopted

3 for the silty clay layer – Model 1: Unit weight = 16 kN/m ; ctop = 6.4 kPa; cmax = 25 kPa; cratio =

3 1.25kPa/m; Model 2: Unit weight = 16 kN/m ; ctop = 10 kPa; cmax = 25 kPa; cratio = 0.87 kPa/m. The geotechnical model associated with the results of the site investigation is indicated in Figure 7 by “Site Investigation Model.”

Figure 7: Back analyses using three geotechnical models.

As mentioned previously, there is widespread agreement that river drawdown adversely affects riverbank stability (e.g. Springer, 1981; Mayo, 1982; Thorne, 1982; Springer et al., 1985; Dahm et al., 1988; Arnott, 1994; Budhu and Gobin, 1995). In general, there are two types of drawdown: short- and long-term, both of which remove the lateral support from the channel and destabilise the bank. The difference being, in short-term drawdown, the pore water pressures within the bank do not have time to equilibrate with the drawn-down river level (Budhu and Gobin, 1995). The circumstances at Long Island Marina suggest that long-term drawdown is more relevant, which leads to permanent changes in the properties and behaviour of bank assets and soils (SKM 2010).

With reference to the river level records at Murray Bridge during 2011, the river fluctuated between 0.5 and 1.2 m AHD, and rarely drew down below 0.5 m AHD. Based on this, a reasonable assumption for a low river level threshold, in foreseeable future, is expected to lie between 0 and 0.5 m AHD, although river management authorities are likely to seek to maintain the river at a higher level than this. In order to examine the susceptibility of the riverbank to collapse, simulations were performed at two river levels (0 and 0.5 m AHD) with two geotechnical models (Figure 8).

With reference to the River Murray, a long term FoS of 1.5 has been adopted as the minimum acceptable FoS, accounting for human safety and the protection of assets (DFW, 2010). As a consequence, FoS ≤ 1 is adopted as “Unstable”, which indicates additional stabilisation is needed for stability; 1 < FoS ≤ 1.25 is assumed to be “Quasi stable”, indicating that minor destabilising factors can

Page 17 of 77 Riverbank Collapse Task 5 Report cause instability, such as additional surface loading, severe boat wash, river drawdown and high winds affecting trees; 1.25 < FOS ≤ 1.5 is adopted as “Moderately stable”, implying that moderate destabilising factors can cause instability, such as construction and extensive river drawdown; and FOS > 1.5 was adopted as “Stable”, denoting that only major destabilising factors are likely to cause instability.

2.3 Results The results of the back-analysis at 0 m AHD river level are shown in Figure 8(a). It can be seen that Cross-sections 5, 13 and 21 are about to collapse; 6, 7, 19 and 22 are at the limit of failure; 9, 10, 11, 12, 15, 18, 20 and 23 are moderately stable; and 8, 14, 16, 17, 24 and 25 are stable.

When the river levels are maintained at 0.5 m AHD, as indicated in Figure 8(b), it is observed that Cross-section 13 is about to collapse; Cross-sections 5, 6, 7, 19, 21 and 22 are at the limit of failure; 9, 10, 11, 12, 20 and 23 are moderately stable; and 8, 14, 15, 16, 17, 18, 24 and 25 are stable.

(a)

(b)

Figure 8: Factors of Safety of neighbouring cross sections (0 and 0.5 m AHD).

Page 18 of 77 Riverbank Collapse Task 5 Report

As shown in Figure 9, the risk of riverbank instability is, in general, reduced by increasing river levels. Hence, when the river level rises to 0.5 m AHD, a relatively large proportion of the banks examined present as quasi stable rather than moderately stable.

Figure 9: Predictions of riverbank susceptibility with river levels at (a) 0 m AHD and (b) 0.5 m AHD.

The low river levels in the Lower River Murray during the Millennium Drought have caused the banks to remain in a long-term, metastable condition since 2007. The study site was adjacent to Long Island Marina at Murray Bridge in South Australia where a series of riverbank collapses have occurred relatively recently. With high-resolution LIDAR (light detecting and ranging) data and aerial photos, this study has vectorised the collapsed riverbank sections by visual interpretation and 2D back- analyses have been performed on 5 cross-sections in the collapsed region using GIS, site investigation data and the slope stability program, SVSlope. In addition, a series of sensitivity analyses has been conducted at two different water levels: 0 and 0.5 m AHD at 21 cross-sections using the results of back-analyses.

2.4 Summary The results of this work have shown that, when the river level is low, a section of the riverbank is close to collapse, whereas a number of cross-sections adjacent to the study site are susceptible to riverbank collapse and require further investigation. In addition, it has been observed that increased river levels generally stabilise the riverbanks but to a limited extent. Several remedial works may need to be conducted when the river level is predicted to fall for an extended period. Later sections will incorporate climatic factors into the riverbank stability assessment at this and several other sites along the Lower River Murray.

Page 19 of 77 Riverbank Collapse Task 5 Report

3 Back Analysis of Historical Riverbank Collapses

3.1 Introduction For the purpose of understanding the underlying mechanics of the riverbank collapses along the lower reaches of the River Murray, as well as quantifying the effects of river level fluctuation, rainfall and evaporation on riverbank stability, back-analyses have been undertaken to model 4 historical and representative riverbank collapses. These riverbank collapses took place at 4 high-risk sites identified by Sinclair Knight Merz (SKM, 2010a). These sites, which include Mannum (MN); Woodlane Reserve (WR); River Front Road, Murray Bridge (RFR); and White Sands (WS), are associated with 50 of the total 162 reported riverbank collapse-related incidents, according to the inventory of the South Australian Department of Environment, Water and Natural Resources (DEWNR) (formerly DFW) between the years 2007 and 2010 (WaterConnect, 2015). The work reported in this section was recently published by Liang et al. (2015a). The geographic locations and nature of these historical riverbank collapse-related incidents associated with these four sites are summarised in Table 1 and Figure 10.

Table 1: Nature of the historical riverbank collapse-related incidents at 4 study sites.

Sites Bank collapse Bank cracking Tree leaning and collapse Levee problem MN 6 7 7 0 WR 3 3 0 0 RFR 8 3 6 2 WS 3 0 3 1

At MN, a section of the riverbank slumped into the river and some areas at MN were observed to exhibit cracking along the riverbank, suggesting impending failure zones had developed. At WR, a 45 × 14 m section of riverbank collapsed, destroying a pumping station. At RFR, which was discussed in the previous section, an unprecedented period of dry conditions and low flows induced a significant section of riverbank (60 × 20 m, 70,000 m3) to collapse into the river, taking with it three unoccupied vehicles and several trees. At WS, two large riverbank collapses were reported: 20 × 6 m on 14 February 2009 and 25 × 4 m on 22 April 2009 (SKM 2010a).

Stability failures are often back-analysed, as mentioned earlier, to estimate the in situ shear strength at the time of the collapse (Abramson et al. 2002). An alternative approach to back-analysis is the determination of soil shear strength by means of undertaking laboratory and/or in situ testing. However, both laboratory and in situ testing present limitations. Laboratory testing, for example, is associated with a number of shortcomings, such as the sample disturbance, in which the structure of the soil has been substantially disturbed due to sampling, such that subsequent laboratory tests are not representative of in situ conditions. Moreover, to estimate accurately the shear strength of soil materials, the field conditions need to be accurately replicated in the laboratory, including the

Page 20 of 77 Riverbank Collapse Task 5 Report following states prior to failure: effective normal stress acting on the failure surface; pre-existing shear deformation; and the drainage conditions during shear (Tang et al. 1999).

(MN)

(WR)

(RFR)

(WS)

Figure 10: The geographical locations of the four studied sites and major riverbank collapses in the past, as well as the proximities of the borehole and river level observation stations.

Page 21 of 77 Riverbank Collapse Task 5 Report

On the other hand, the examples of the limitations of in situ testing include the uncertainty of the models used to translate the field measurements into their associated soil properties, as well as the human factors in operating the testing equipment and conforming to the standardised test procedures. For these reasons, back-calculated soil shear strengths have advantages over laboratory- determined data, because they represent a large scale of soil mass over slip surfaces and they were determined from an in situ state (Gilbert et al., 1998). A back-analysis assumes: (i) a slope geometry prior to failure; and (ii) a factor of safety (FoS) is equal to unity that is the condition of the failure to have occurred according to the limit equilibrium method to estimate the in situ shear strength at the time of failure (Okui et al., 1997; Gilbert et al., 1998; Urgeles et al., 2006; Bozzano et al., 2012; Harris, Orense & Itoh, 2012; Wang et al. 2013).

3.2 Methodology In this section, an efficient framework for the back-analysis of riverbank collapses is adopted at the 4 aforementioned sites. The framework employs a finite element analysis-based transient water model to evaluate the dynamic distribution of pore water pressure under different circumstances of rainfall, evaporation and river level fluctuations. Subsequently, limit equilibrium slope stability analyses are performed to back-analyse the failure. The work presented in this section aims to: (i) examine the 4 aforementioned riverbank collapses using 2D cross-sectional models; and (ii) determine the soil shear strength profiles from the back-analyses.

In order to determine the location of the collapses with a relatively high degree of accuracy, visual interpretation of high-resolution aerial images has been implemented in conjunction with DEWNR’s inventory of historical riverbank failures (shown as green squares in Figure 11). For each riverbank collapse, two aerial photographs have been used to identify the extent of the collapse regions. Figure 11(a), (c), (e) and (g) were acquired in March 2008 with a 0.5 metre resolution; whereas Figure 11(b), (d), (f) and (h) incorporate a 0.2 metre resolution which were acquired after the recorded collapses (May 2010). As shown in Figure 11, the examined regions of collapse have been vectorised with dotted areas within ArcGIS, and have been linked to the closest DEWNR collapse record (incident IDs 23 (MN), 47 (RFR), 108 (WS), 113 to 115 (WR) in the Riverbank Collapse Hazard Incident Register [WaterConnect, 2015]).

A transient, unsaturated flow-based riverbank stability model has been developed in two dimensions to facilitate back-analyses at the 4 aforementioned sites along the lower reaches of the river. The model incorporates riverbank geometry, geotechnical properties, river level variation, and rainfall and evaporation. Back-analyses have been performed to obtain the closest match between the predicted and actual date of failure, while comparing the predicted failure geometry with the high-resolution aerial images.

Page 22 of 77 Riverbank Collapse Task 5 Report

(a) (b)

(c) (d)

(e) (f)

(g) (h)

Figure 11: Adopted visual interpretation method of high-resolution aerial images: (a), (c), (e) and (g) are aerial photographs acquired in March 2008 at MN, WR, RFR and WS, respectively; (b), (d), (f), and (h) are aerial photographs acquired in May 2010 at MN, WR, RFR and WS, respectively.

Page 23 of 77 Riverbank Collapse Task 5 Report

An elevation comparison method has been employed at each of the sites to assist with visual interpretation. Specifically, LIDAR (light detection and ranging) DEMs obtained in 2008 [Figure 12(a)] were used to verify the elevation of the examined collapses with the LIDAR DEMs obtained in 2010 [Figure 12(b)], which include the elevation of the river level. For example, point A in Figure 12(a), where the elevation of the bank was at 1.484 m AHD prior to the collapse, is compared with point B, where the river level was at –0.45 m AHD subsequent to the failure. By comparing the DEMs, each of the examined regions [Figure 11(b), (d), (f) and (h)] is confirmed as collapsed rather than simply submerged beneath a higher river level.

(a) (b) Figure 12: Example of adopted elevation comparison method on DEMs at WR (a) 1 m resolution DEM acquired in 2008; (b) 0.2 m resolution DEM acquired in 2010).

With reference to the DEWNR riverbank collapse inventory (WaterConnect, 2015), visual interpretation was implemented on high-resolution aerial images to examine the region of collapse within ArcGIS. The 2D cross-sections that coincided with the collapse regions have been used to analyse the evolution of the FoS of the riverbank at the 4 sites. The groundwater flows in the saturated and unsaturated zones within the riverbank have been simulated using the finite element method within the PC-based program SVFlux (SoilVision 2009a). The 2D limit equilibrium slope stability calculations have again been performed using SVSlope (SoilVision 2009b). Specifically, in SVFlux, the variations in rainfall, evaporation and river level fluctuations are entered to model the dynamic distribution of pore water pressure within the riverbank. These are subsequently imported into SVSlope to assess the stability of the riverbank against time. By considering the vertical extents of the collapses, back-analyses have been performed on the layer of high plasticity Silty Clay (CH), as outlined below, using the Unsaturated Fredlund soil model (Fredlund et al. 1996) informed by geotechnical investigations performed near each site (Jaksa et al., 2015).

In this section, the geometries of the riverbank associated with the 4 sites have been obtained from DEMs using ArcGIS. As shown in Figure 12, each of the cross-sections initially comprised of 401 points,

Page 24 of 77 Riverbank Collapse Task 5 Report with a 10 m lateral interval between each neighbouring point. Again, based on the bilinear interpolation method, the elevation data have then extracted from the two DEMs (a 1 m resolution DEM acquired in 2008 and a 0.2 m resolution DEM acquired in 2010, as mentioned above) and assigned to those 401 points for each cross-section, with singular points avoided. As indicated in Figure 11, the locations of the 4 cross-sections coincide with and cross each examined region of collapse. This process generated high-resolution cross-sections, which provide effective modelling of the riverbanks in both SVFlux and SVSlope.

In October and November 2009, geotechnical investigations were performed by SKM at three sites, including MN (3 boreholes and 1 piezocone [CPTu]), WR (2 boreholes and 5 CPTus) and RFR (2 boreholes and 13 CPTus) (SKM 2010b). Based on the data from investigations, summarised in Table 2 and Figure 4, the geotechnical models have been developed at each of the 4 sites. The WS and RFR sites share a common geotechnical model, which was obtained from the RFR geotechnical investigation, whereas the models for MN and WR have been established from their respective investigations, as indicated above.

For the purpose of modelling the variation of soil suction under saturated and unsaturated conditions, soil water characteristic curves (SWCCs) are used in each of the soil layers within SVFlux. The SWCC parameters are estimated from the particle size distributions (PSDs) in each soil layer using the Zapata method (Torres Hernandez, 2011), which incorporates the weighted Plasticity Index (wPI). Adopting Equation (1), wPI is determined based on the PSDs from the borehole soil samples (Table 2).

wPI = IP × P200⁄100 (1)

where IP is the plasticity index and P200 is the percentage of soil passing size #200 (US sieve size #200, which is equivalent to a 0.074 mm aperture).

The saturated hydraulic conductivity, Ksat, is derived from a CPTu pore pressure dissipation test performed at 5 m depth in Murray Bridge, as shown in Figure 13(b). Based on relationships between heavily consolidated silts and clays, monotonically decreasing excess pore pressure with time response and the saturated hydraulic conductivity (Burns and Mayne 1998) for the CH and CL layers,

–5 Ksat is found to equal approximately 9.9 × 10 m/day, as shown in Table 2 and Figure 13(b). In unsaturated soil, the unsaturated hydraulic conductivity, K, varies with respect to matric suction, and is calculated indirectly from the Fredlund and Xing (1994) estimation associated with Ksat.

The river level data used in this study are obtained from 4 observation stations: A4261161, A4260547 (WaterConnect 2014), A4261162 and A4260522 (MDBA 2014), as shown previously in Figure 10. The climatic data, which include mean daily rainfall, mean daily temperature, evaporation and humidity,

Page 25 of 77 Riverbank Collapse Task 5 Report at the 4 sites are collected from the Bureau of Meteorology (2014). In order to back-analyse the known riverbank collapses under the combined influence of river level fluctuation, rainfall and evaporation, a one-month historical record was adopted prior to the recorded date of each riverbank collapse.

As shown in Figure 14, unexpected low inflows, combined with evaporation of the lower lakes, resulted in the daily river levels remaining between –1.1 to –0.8 m AHD at these 4 aforementioned sites. More specifically, compared with the relatively constant daily river levels at WR and RFR, significant high flow events were observed at MN and WS in the last 7 days of the month. Approximately 30 mm and 35 mm rainfall were observed at MN and WS, respectively, however, the rainfall at WR and RFR sites was negligible.

Table 2: Soil properties for saturated and unsaturated flow modelling.

Elevation K θ ρ w I γ P Layer sat sat P 200 (m AHD) (m/day) (%) (t/m3) (%) (%) (kN/m3) (%) Mannum (MN) Silty/Clayey Sand 3.5 to 1 13.51 45.2 1.7 17.1 2 20 ± 1 21 (SM/SC)

Silty Clay (CH) 1 to –1 9.9 x 10–5 59.8 1.4 26.6 16 17 ± 1 20

Silty Clay (CH) –1 to –5 9.9 x 10–5 80.7 0.95 79 42 17 ± 1 15

Clayey Sand (SC) > –5 0.187 49.7 1.6 59 38 20 ± 1 20 Woodlane Reserve (WR) Sand (SP) 2.5 to 2 15.21 62 1.35 35.4 40 17 ± 1 78

Silty Clay (CH) 2 to –1 9.9 x 10–5 76 1 56 41 17 ± 1 37

Clayey/Silty Sand –1 to –5 0.187 57 1.7 48.6 17 20 ± 1 29 (SC/SM)

Sandy/Silty Clay (CL) –5 to –11 9.9 x 10–5 52 1.6 22 14 20 ± 1 68

Silty Sand/ Gravel > –11 0.187 51 1.6 22 3 20 ± 1 18 (SM) River Front Road (RFR) and White Sands (WS) Silty/Clayey Sand 1 to 0 13.51 52.7 1.25 42 30 18 ± 1 35 (SM/SC)

Silty Clay (CH) 0 to –20 9.9 x 10–5 63.2 1.01 61 50 16 ± 1 67

Clayey Sand/Sandy > –20 0.187 54.6 1.23 44 50 17 ± 1 59 Clay (SC/CL)

Page 26 of 77 Riverbank Collapse Task 5 Report

(a)

(b) Figure 13: Results of CPTu sounding (a) profile; and (b) pore pressure dissipation test results.

Page 27 of 77 Riverbank Collapse Task 5 Report

(a) (b)

(c) (d) Figure 14: Daily river levels and daily rainfall recorded at (a) Mannum (MN) site in April 2009; (b) Woodlane Reserve (WR) site in February 2009; (c) River Front Road, Murray Bridge (RFR) site in February 2009; and (d) White Sands (WS) site in April 2009.

The dates of each riverbank collapse incident have been obtained from the most relevant DEWNR record, and the back-analyses are benchmarked against these dates. The riverbank profiles at the 4 sites were modelled previously by SKM (2010a) and Coffey (2012) using separate soil layers informed by borehole logs from SKM (2010a, 2010b, 2010c, 2010d, 2010e). Specifically, the static SKM and Coffey analyses, as summarised in Table 3 and Figure 15 to Figure 18, incorporated effective stress parameters for the Fill and SC layers using the Mohr-Coulomb failure criterion, while total stress parameters were adopted in the CH layers using the following depth-dependent, undrained model: linearly-increasing cohesion with depth, with ctop quantifying the cohesion (kPa) at the upper layer boundary, cratio representing the gradient of increasing cohesion with depth and capped at a maximum value of cmax. Each of the values recommended by SKM (2010a) and Coffey (2012) were inferred from laboratory and in situ test results obtained from their respective geotechnical investigations. However, in order to accommodate the effects of positive and negative pore water pressures in the partially-saturated riverbanks, an unsaturated, effective stress analysis, based on the Unsaturated Fredlund model (Fredlund et al. 1996) is performed on the Fill and CH layers in the present paper, as shown by Equation (2).

′ ′ 푏 휏 = 푐 + (휎푛 − 푢푎)푡푎푛∅ + (푢푎 − 푢푤)[휃(푢푎 − 푢푤)⁄휃푠] 푡푎푛∅ (2)

Page 28 of 77 Riverbank Collapse Task 5 Report

where 휃(푢푎 − 푢푤) = the volumetric water content at any suction; 휃푠 = saturated volumetric water content; b = a fitting parameter that has a value close to unity for sands and increases with plasticity. At each site, the geotechnical properties of the high plasticity, clay (CH) layers are varied marginally, with respect to the results of the corresponding site investigations, until the FoSs of unity were obtained in close proximity to the recorded dates of collapse. As shown in Figure 15 to Figure 18, values of FoS = 1.0 were obtained from the most appropriate Unsaturated Fredlund soil model as summarised in Table 3. For example, at the River Front Road (RFR) site at Murray Bridge, the soil model for the CH layer adopted c' = 0 kPa; φ' = 22°, which aligns reasonably well with the single CU triaxial test performed by SKM (2010b) (φ' = 27°).

Table 3: Geotechnical models of the clay layer obtained from back analyses.

Effective stress Undrained parameters for CH Undrained parameters for CH Sites parameters from back layer from SKM (2012) layer from Coffey (2012) analysis

cu-B = 50 ± 10 (kPa) MN cu-B1 = 17.5 ± 2.5 (kPa) c' =0, '=23° cu-B2 = 14± 2 (kPa) cu-top = 5.5 (kPa)

WR cu = 20 ± 5 (kPa) cu-ratio = 1.25 (kPa/m) c' =0, '=20°

cu-top = 10 ± 5 (kPa) cu-max = 25 ± 5 (kPa) RFR cu-ratio = 1.25 (kPa/m) c' =0, '=19° cu-max = 25 ± 5 (kPa)

WS N/A N/A c' =0, '=24°

3.3 Results As mentioned above, the models are validated in two ways. Firstly, the predicted dates of bank collapse are compared with the historical collapse dates from DEWNR inventory. As can be seen from Table 4 and Figure 19 to Figure 22 the predicted dates compare very favourably with the historical dates. This is not particularly unexpected, given that the values of φ' were varied until the predicted and historical collapse dates were in good agreement. However, Figure 19 to Figure 22 are particularly encouraging in that the FoS time series, prior to collapse, consistently plot above unity, which is consistent with increasing pore water pressure prior to failure. In addition, the second independent validation measure is the volume of the collapsed region. By scrutinising the high- resolution aerial images (Figure 11), the dimensions of the collapsed regions can be determined, as explained earlier. Table 4 also shows the predicted widths of the collapsed regions from the back- analyses of the 4 sites. As can be seen, the predicted widths compare extremely well with those of the actual collapses.

Page 29 of 77 Riverbank Collapse Task 5 Report

Figure 15: Riverbank stability analysis of the Mannum (MN) site on 21 April 2009.

Figure 16: Riverbank stability analysis of the Woodlane Reserve (WR) site on 26 February 2009.

Figure 17: Riverbank stability analysis of the River Front Road, Murray Bridge (RFR) site on 6 February 2009.

Page 30 of 77 Riverbank Collapse Task 5 Report

Figure 18: Riverbank stability analysis of the White Sands (WS) site on 23 April 2009.

Table 4: Model validation.

Width of collapsed Recorded date of Predicted date of Predicted width of Sites region from aerial collapse collapse collapsed region photos MN Approx. Day 20 Day 23 8 m 8 m

WR Approx. Day 28 Day 26 18 m 17.7 m

RFR Approx. Day 4 Day 6 20 m 19.5 m

WS Approx. Day 22 Day 23 6 m 5 m

Figure 19: Riverbank collapse factor of safety time series for Mannum (MN) in April 2009.

Page 31 of 77 Riverbank Collapse Task 5 Report

Figure 20: Riverbank collapse factor of safety time series for Woodlane Reserve (WR) in February 2009.

Figure 21: Riverbank collapse factor of safety time series for River Front Road (RFR) in February 2009.

Figure 22: Riverbank collapse factor of safety time series for White Sands (WS) in April 2009.

Page 32 of 77 Riverbank Collapse Task 5 Report

3.4 Summary In summary, back-analysis of riverbank collapse has been undertaken and applied to 4 high-risk sites along the lower reaches of the River Murray. A GIS framework, incorporating LIDAR digital elevation models and high-resolution aerial images, has been used to determine the riverbank geometries prior collapse and cross-examine the actual collapsed regions from the inventory. Geotechnical data, which have been used in the back-analyses, were obtained from the site investigations performed at the 4 sites. A transient, unsaturated flow-based riverbank stability model was implemented using SVSlope in conjunction with SVFlux.

The study has shown that: (a) The results of back-analysed soil shear strengths at the 4 sites have shown very good consistency with those proposed by the geotechnical consultant (SKM) commissioned to undertake site investigations adjacent to the collapse sites; (b) Model validation demonstrates the adopted framework provides reliable riverbank stability predictions; and (c) The integration of GIS with high-resolution spatial data facilitates the process of collapsed- region identification, model geometry development and the calculation of the dimensions of the collapsed regions.

Page 33 of 77 Riverbank Collapse Task 5 Report

4 Effects of River Level Fluctuation and Climate on Riverbank Stability

This section presents the results of numerical modelling performed to examine the influence of river level fluctuations and climatic factors on the occurrence of riverbank collapses. This work was originally published by Liang et al. (2015c) but has been updated to include recent developments.

4.1 Background The effects of climate and river level fluctuation on the stability of riverbanks have been extensively explored and discussed by several researchers (Hooke, 1979; Springer, 1981; Thorne, 1982; Osman and Thorne, 1988; Thorne and Osman, 1988; Casagli et al., 1999; Green, 1999; Pauls et al., 1999; Rinaldi and Casagli, 1999; Lane and Griffith, 2000; Zhang et al., 2005; Jia, 2009; Li et al., 2011). Compared with the assessment of landslides in mountainous regions, changes in pore water pressure, in particular soil suction (negative pore water pressure), plays a more fundamental role in the stability of riverbanks (Hooke, 1979; Thorne, 1982; Casagli et al., 1999; Rinaldi and Casagli, 1999; Abramson et al., 2002). The pore water pressure significantly affects riverbank stability by changing the soil shear strength (Section 2). As stated by Fredlund (2006), soil suction can vary from zero to approximately 1,000 MPa in the unsaturated (or vadose) zone (Bouwer, 1978) and is highly dependent on the properties of the soil, hydrological conditions and soil–atmospheric interactions (Lane and Griffith, 2000).

Rainfall has long been recognised as one of the most significant factors responsible for initiating slope failures in many tropical or subtropical regions (Zhang et al., 2011). Generally, rainfall-induced slope failures are observed as shallow failures; however deep-seated rotational failures are also reported (Zhang et al., 2011). According to Rahardjo et al. (2007), the initial FoS of the slope is determined by the slope geometry and initial water table while the actual rainfall-induced failure conditions are greatly influenced by rainfall characteristics and soil properties. The nature of the rainfall, such as intensity, duration, spatial distribution and antecedent characteristics, significantly influences the occurrence of rainfall-induced landslides by affecting the pore water pressure distribution and increases the self-weight of the slope materials (Abramson et al., 2002; Rahardjo et al., 2008; Rahimi et al., 2011; Zhang et al., 2011; Mukhlisin and Taha, 2012). Based on several studies performed in Asia, Lumb (1975) and Brand (1984) indicated that antecedent rainfall has a negligible effect on local rainfall-induced landslides. They concluded that rainfall intensity and duration had the most profound influence on the slope failure due to localised high conductivity soils. Later, Rahardjo et al. (2001, 2008) and Rahimi et al. (2011) undertook a series of more detailed studies along the Sieve River in Italy in relation to rainfall associated with soils of different conductivities. Their work showed that the rate of decrease in FoS, the time corresponding to the minimum FoS and the value of the

Page 34 of 77 Riverbank Collapse Task 5 Report minimum FoS are all controlled by the rainfall distribution. In comparison, rainfall can induce up to a 45% reduction in the FoS of soil slopes with lower fines content and higher saturated permeability

-5 -6 (Ksat ≥ 10 m/s) than those with a high fines content and low permeability (Ksat ≤ 10 m/s) (Rahimi et al. 2011). These geotechnical properties greatly influence the behaviour of rainfall-triggered slope failures because they affect rainwater seepage and infiltration. Rahardjo et al. (2007) examined the relationship between Ksat, soil suction, FoS and the magnitude of rainfall. Their work found that,

-6 under modest rainfall intensity (10 mm/h), soils with Ksat = 10 m/s were associated with the lowest

-5 FoS, followed by soil with Ksat = 10 m/s. In contrast, under relatively intense rainfall (greater than

-5 200 mm/h), soils with Ksat = 10 m/s were associated with the lowest FoS.

River level fluctuation has been shown to influence riverbank stability in two important ways: (i) its effect on reducing negative pore water pressure and, hence, its consequent reduction in soil strength, and (ii) the hydrostatic pressure it applies to stabilizing the riverbank (Casagli et al., 1999; Green, 1999; Rinaldi and Casagli, 1999). Due to the limited models available at the time, studies in the 1980s typically proposed simple hypotheses on pore water pressure conditions of the riverbank (dry or total saturated) and adopted relatively simple solutions for slab failures (Hooke, 1979; Higgins, 1980; Springer, 1981; Thorne, 1982). Later in the 1990s and 2000s, with developments in unsaturated soil mechanics theory, the effect of pore water pressure on unsaturated riverbanks and confining pressure became more widely accepted and included into drawdown analysis and research of riverbank stability (Casagli et al., 1999; Green, 1999; Pauls et al., 1999; Rinaldi and Casagli, 1999; Lane and Griffiths, 2000; Rinaldi et al., 2004; Zhang et al., 2005; Berilgen, 2007; Jia et al., 2009; Li et al., 2011; Yang et al., 2010). It is generally accepted that when rivers experience an initial high-flow period, the riverbanks are stable due to the supportive effect of the hydrostatic pressure of the water. However, the processes of erosion and soil saturation during high flow events weaken many parts of the bank by undermining it and reducing the effective strength, respectively (Twidale, 1964; Hooke, 1979; Springer, 1981; Thorne, 1982; Thorne et al., 1997). Berilgen (2007) indicated that the stability of a submerged slope during drawdown greatly depends on the rate of pore water drainage. While during initial low-flow periods, the matric suction (the suction due to capillary action and water surface tension) occasionally allows the riverbank remain stable at steep angles (Casagli et al., 1999). However, subsequent rainfall increases the dead weight of the bank material and reduces the matric suction which might be sufficient trigger a mass failure (Rinaldi et al., 2004).

A very limited number of studies modelled the coupling of climatic factors and river level fluctuations. Casagli et al. (1999) used tensionmeters, piezometers and a rain gauge on the Sieve River to monitor the matric suction evolution and riverbank stability in a semi-arid climate with daily river flow changes over a 16-month period. Later, based on these monitoring works, transient modelling of a drawdown failure, which occurred on December 14, 1996, was carried out by Rinaldi et al. (1999). In the transient model, the research period (December 13–18, 1996) was divided into 24 time steps to

Page 35 of 77 Riverbank Collapse Task 5 Report examine the behaviour of the riverbank under different rainfall and flow events. It was indicated that the minimum FoS always occurred after the peak level of the Sieve River, and no later than 5.5 hours after. The result suggests that riverbank collapse on Sieve River is dominated by river level fluctuations, primarily due to the reduction in the stabilizing influence of hydrostatic pressures due to river level drawdown, and marginally due to rainfall. This conclusion is in agreement with the earlier work of Twidale (1964), Springer (1981), Thorne (1982, 1988), Thorne et al. (1997) and others.

The section aims to: (i) model the riverbank collapse incident, in both 2D and 3D, which occurred at Long Island Marina, Murray Bridge, South Australia, on February 4, 2009; (ii) examine the influence and sensitivity of river level fluctuations and climatic factors on riverbank stability; and (iii) determine the dominant triggers affecting collapse. Specifically, the study implements a transient water model in both two and three dimensions to evaluate the FoS of the riverbank by determining the temporal distribution of pore water pressure in the riverbank and simulating its dynamic response to river level fluctuations and rainfall. Due to the relatively low width-to-height ratio of the riverbank collapsed region, as indicated by bathymetric surveys, a 3D analysis is also performed to complement and validate the date and extent of the collapse and to compare this with the results derived from the 2D model. Details of the Long Island Marina site are presented in the following section.

4.2 Study Area The study area is located along the lower River Murray adjacent to the Long Island Marina, Murray Bridge, South Australia, as shown in Figure 23(a). From the DEWNR inventory, several incidents were recorded at Long Island Marina between 2008 and 2010. These included 5 major riverbank collapses, an identified significant tension crack, two riparian trees leaning into the river, and one levee bank related incident (Section 2). On February 4, 2009 a significant riverbank collapse incident occurred which involved a 100 m stretch of riverbank [Region (iii) in Figure 23(b)], which failed without warning. Several large trees, three unoccupied vehicles and an estimated 70,000 m3 of bank material collapsed into the river (SKM 2011). The bathymetry, as shown in Figure 23(b), illustrates the scale of a series of collapses, which occurred in the vicinity of Long Island Marina.

As mentioned earlier, a site investigation was performed in October and November 2009 at the Long Island Marina site by SKM (2010b) in order to obtain geotechnical information related to the collapse. Details of the investigation and geotechnical characteristics are presented later.

Page 36 of 77 Riverbank Collapse Task 5 Report

(a) Location of study area.

(b) Aerial bathymetry image at Long Island Marina, Murray Bridge showing the contours of the riverbed and the collapse of 70,000 m3 of material into the River Murray (Source: Beal et al., 2010).

(c) 2D cross-section showing river level on the date of collapse.

Figure 23: Details of the Long Island Marina site.

Page 37 of 77 Riverbank Collapse Task 5 Report

4.3 Methodology In order to investigate the factors influencing the Long Island Marina riverbank collapse, a transient slope stability model was developed in both 2D and 3D. The model incorporates riverbank geometry, geotechnical properties, river level variation, and rainfall and evaporation. The implementation of each of these into the transient model is outlined in the sub-sections that follow.

SVFlux (SoilVision 2009a) has again been used to undertake finite element analyses of saturated and unsaturated groundwater flows, in conjunction with SVSlope (SoilVision 2009b), which performs limit equilibrium slope stability calculations, in this case, in both 2D and 3D. Specifically, SVFlux has been employed to model the distribution of pore water pressure against time subject to different circumstances of rainfall, evaporation and river level fluctuations. These have subsequently been imported into SVSlope to examine the stability of the riverbank at each time-step. The limit equilibrium method (Bishop’s method of slices) is employed to perform the limit equilibrium slope stability calculations. The sliding surfaces are determined using the grid and tangent and slope limit methods, and for the 3D model, reference is made to the actual dimensions and location of the collapsed region. In order to determine the most appropriate soil properties for the high plasticity, Silty Clay layer, as explained later, and to accommodate variability of soil properties, back-analyses are again performed to obtain the closest match between the predicted and actual date of failure (in the 2D analysis) and the dimensions of the collapsed region (3D analysis).

As mentioned above, in order to determine the riverbank geometries associated with the Long Island Marina collapse, the GIS framework is adopted using topographic information obtained from the DEM; the collapsed regions are examined by visual interpretation of high resolution aerial images; and the dimensions of the simulated collapsed regions are validated against these high-resolution aerial images.

As mentioned previously, the Long Island Marina riverbank failure is examined in both two and three dimensions. It is commonplace in the assessment of slope stability for 2D cross-sections to be adopted (Raharjo et al., 2001, 2008; Rahimi et al. 2011). Stark (2003) undertook a study where he compared several 2D and 3D analyses and concluded that, if the width-to-height ratio of the collapsed region is larger than 6, the difference in the factors of safety obtained from the 2D and 3D analyses is marginal. In the Long Island Marina riverbank collapse, as shown by Region (i) in Figure 23(b) from the bathymetric data and Figure 23(c) from the cross-sectional model, the width-to-height ratio is found to be equal to 5.2 (78 m in width and 15 m in height). Hence, 3D modelling is beneficial in this study by affording an additional analytical technique to validate the volume of the collapsed region and to compare the FoS time history with that obtained from the 2D model.

Page 38 of 77 Riverbank Collapse Task 5 Report

In order to establish accurate riverbank geometries, ArcGIS is again used, incorporating the bilinear interpolation method based on elevation extraction. As shown in Figure 24(a) and (b), the 2D cross- sectional model is initially obtained as a polyline comprised of 401 points with a 10 m interval between neighbouring points. The location of the cross-section used in the 2D analysis is selected to coincide with the recorded collapse, as indicated in the DEWNR riverbank collapse inventory and the bathymetric data [Region (i)], as shown in Figure 23(b) and Figure 24(b). Based on the bilinear interpolation method, the elevation values were extracted from two DEMs (a 1 m resolution model acquired in 2008 and b 0.2 m resolution model acquired in 2010) and assigned to 401 points, with singular points avoided. This approach yields accurate and high-resolution cross-sections, suitable for subsequent importing into SVFlux and SVSlope.

By utilising the high-resolution DEMs, the topography of the riverbank region of interest is extracted using custom masks within ArcGIS, as shown in Figure 24(b) for the purpose of reproducing the riverbank model in three dimensions. Due to the nature of DEM data, the elevation values that are extracted are irregularly distributed. To resolve this issue, the inverse distance weighting method (IDW) is used to interpolate a new raster based on the former irregularly distributed elevation values. As shown in Figure 24(c), the geometry of the riverbank was established in 3D in SVFlux3D from the updated elevation values extracted from the new raster.

The dimensions of the collapsed regions of the riverbank need to be identified with a relatively high level of accuracy in order to facilitate the proposed modelling. In this study, this has been achieved by undertaking a visual interpretation of the high-resolution aerial images taken on different dates (Section 2). By comparing the 0.5 m resolution aerial photos acquired in March 2008 [Figure 25(a)] with the 0.2 m resolution photos acquired in May 2010 [Figure 25(c)], the collapsed Region A has been accurately identified and vectorised. Comparing the collapsed regions (i), (ii) and (iii) shown in Figure 23(b), Region A [shown by the dotted region in Figure 25(c)] is the section above the river level. To assist with the visual interpretation of the collapsed region, LIDAR DEMs obtained in 2008 [Figure 25(b)] and 2010 [Figure 25(d)] have been employed, especially to verify the elevation of the riverbank prior to the collapse with river levels at the same location subsequent to the collapse. For example, location A' [shown by the red dot in Figure 25(c)], where the riverbank was at 0.6 m AHD prior to the collapse, is compared with the river level at location B', which is at −0.45 m AHD subsequent to the failure. By adopting this comparative process, the identified region A is confirmed as having collapsed rather than simply having been submerged.

The geotechnical model is developed from the results of a geotechnical investigation performed at Long Island Marina by SKM (2010b), which included the drilling of two boreholes to a depth of 20 m, 12 piezocone tests (CPTUs) with pore water pressure (u) measurements, and a series of vane shear and laboratory tests (Figure 26 to Figure 29).

Page 39 of 77 Riverbank Collapse Task 5 Report

(a) Elevation extraction from the 2008 (b) Elevation extraction from the 2010 LIDAR LIDAR DEM. DEM, with river level extracted.

(c) 3D topographical model. Figure 24: Riverbank geometry definition.

In accordance with the borehole logs from SR-BH1 and SR-BH2 (SKM 2010b), the bank profile at Long Island Marina has previously been modelled by SKM (2010a) and Coffey (2012) using three separate soil layers. Specifically, in the SKM and Coffey models, as summarised in Table 5, an effective stress analysis was performed in the Fill and SC layer using the Mohr-Coulomb failure criterion, while a total stress analysis was adopted in the CH layer using a depth-dependent, undrained soil model. The adopted depth-dependent, undrained soil model incorporates linearly-increasing cohesion with depth, with ctop quantifying the cohesion (kPa) at the upper layer boundary, cratio representing the gradient of

Page 40 of 77 Riverbank Collapse Task 5 Report

(a) Aerial photo in 2008. (b) LIDAR DEM before collapse.

(c) Aerial photo in 2010 with collapsed region (d) LIDAR DEM after collapse. highlighted. Figure 25: Example of adopted visual interpretation process on high-resolution, aerial images within ArcGIS.

increasing cohesion with depth and capped at a maximum value of cmax (SKM 2010a). However, in order to accommodate the effects of positive and negative pore water pressures in unsaturated riverbanks, an unsaturated stress analysis is performed in the Fill and CH layer for this paper, which will be discussed later. As summarised in Table 5, the values recommended by SKM (2010a) and Coffey (2012) were inferred from laboratory and in situ test results obtained from their respective geotechnical investigations.

Table 5: Soil parameters for stability assessment.

Undrained parameters for CH Elevation c' γ K θ Layer φ' layer from: sat sat (m AHD) (kPa) (kN/m3) (m/day) (%) SKM (kPa) Coffey (kPa) Silty/Clayey Sand 1 to 0 c'=2 ± 2 28° ± 2° 18 ± 1 – – 13.51 52.7 (SM/SC)

c = 10 ± 5 c = 5.5 Silty Clay u-top u-top 0 to –20 0 27° 16 ± 1 c = 1.25 c = 1.25 9.9x10–5 63.2 (CH) u-ratio u-ratio cu-max = 25 ± 5 cu-max = 25 ± 5

Clayey Sand/ Sandy > –20 c'=2 ± 2 30° ± 2° 17 ± 1 – – 0.187 54.6 Clay (SC/CL)

Page 41 of 77 Riverbank Collapse Task 5 Report

(a) Moisture content and plasticity index profile, (b) Dry density profile and undrained shear Type-A represents SM/SC, Type-B represents CH strength from CPTu. and Type-C represents SC/CL. Figure 26: Geotechnical profiles based on soil samples taken from SR-BH1 and SR-CPTu6s at Long Island Marina.

Figure 27: Particle size distributions based on the soil samples from four different depths in borehole SR-BH1.

Page 42 of 77 Riverbank Collapse Task 5 Report

In order to determine accurately the variation of total soil suction with time and moisture, a soil water characteristic curve (SWCC) is adopted in the finite element method analyses for unsaturated soils within SVFlux. The Zapata et al. (2000) method is used to estimate the SWCC in each of the soil layers from the particle size distributions (PSDs) shown in Figure 27. The weighted Plasticity Index (wPI) is used within the Zapata method and is defined as an indicator for water adsorption and retention in soil particles with different surface areas (Zapata et al. 2000). As stated by Zapata at al. (2000), a proportional relationship exists between the equilibrium soil suction at a given degree of saturation and the specific surface area of the soil. Compared with the plasticity index (IP), wPI is considered as a better indicator, especially for the soils that have a high IP but are associated with relatively small surface areas (Zapata et al., 2000). As shown in Table 6, for soil with a high plasticity

(wPI > 0), such as the Silty Clay (CH), the Fredlund et al. (1995) parameters af, bf, cf and hf are calculated based on the relationships shown in Table 6 to estimate the SWCC for each of the three layers (Figure 28). In contrast, for granular soils with wPI = 0 (i.e. non-plastic soils), the parameter D60, which represents the grain diameter corresponding to 60% passing by weight from the PSD, is employed.

Figure 28: Estimated SWCCs for the three soil layers at Long Island Marina using the Fredlund and Xing (1994) estimation method.

The saturated hydraulic conductivity, Ksat, is derived from a CPTu pore pressure dissipation test performed in the CH layer at 5 m depth at the Murray Bridge site [Figure 13(b)]. Based on relationships between measured excess pore pressures and the saturated hydraulic conductivity

–5 (Burns and Mayne, 1998) for the clay layer, Ksat is found to equal approximately 9.86 × 10 m/day, as shown in Table 5. The values of Ksat for the Fill and SC layers are obtained from literature (Table 5) as no measurements are available. In unsaturated soil, the unsaturated hydraulic conductivity, K, varies with respect to matric suction, and is calculated indirectly from the Fredlund and Xing estimation

Page 43 of 77 Riverbank Collapse Task 5 Report

Table 6: Equations used to calculate Fredlund and Xing (1994) SWCC fitting parameters based on the soil grain size distribution.

Soil with high plasticity Non-plastic soil

푤푃퐼 > 0 푤푃퐼 = 0

푎푓 = 1.14푎 − 0.5 푎푓 = 32.825 푙푛(푤푃퐼) + 32.438 −6 4.34 푎 = −2.79 − 14.1 푙표𝑔(퐷20) − 1.9 ∗ 10 푃200 + 7 푙표𝑔(퐷30) + 0.055퐷100

푏푓 = 0.936푏 − 3.8

−0.3185 퐷90 0.57 푏푓 = 1.42(푤푃퐼) 푏 = (5.39 − 0.29 푙푛 (푃200 ⁄ ) + 3퐷0 퐷10 1.19 30 0.1 + 0.0021푃200 )( ⁄ ) 푙표𝑔(퐷90) − 푙표𝑔(퐷60)

0.758푐 푐푓 = 0.26푒 + 1.4퐷10 푐푓 = −0.2154 푙푛(푤푃퐼) + 0.7145 푐 = 푙표𝑔(20⁄ )1.15 − 1 + 1⁄ 푙표𝑔(퐷30) − 푙표𝑔(퐷10) 푏푓

ℎ푟 = 500 ℎ푟 = 100

4(푙표푔(퐷90)−푙표푔(퐷60)) ⁄ +푙표푔(퐷60) 퐷100 = 10 3 푤푃퐼 = 퐼푃 × 푃200⁄100 −3(푙표푔(퐷30)−푙표푔(퐷90)) ⁄ +푙표푔(퐷30) 퐷0 = 10 2

Note: 푃200is the percentage of soil passing the US standard sieve #200. IP is the plasticity index as shown in

Figure 26(a); and 퐷% is the grain diameter related to the percentage of passing in mm.

associated with Ksat (Fredlund et al., 1995). Adopting Equation (3), the saturated volumetric water content, θsat, is determined by the following equation and based on the tests from borehole soil samples (Table 5).

ρ θsat = 1 − ⁄ (3) Gsρw(1 + w)

where ρ is the dry density; ρw is the density of water at 4°C and w is the moisture content.

In order to model the influence of rainfall on riverbank stability it is necessary to estimate the runoff coefficient so that infiltration of rainfall into the riverbank can be evaluated. It is well understood that the runoff coefficient is a function of hydraulic conductivity and thickness of the near-surface soil layers, the type of surface vegetation and land use. Due to the high rate of evapotranspiration, temporal and spatial variability in rainfall intensity and frequency, and the generally flat topography across most of South Australia (National Water Commission, 2005) the runoff coefficient is relatively small. Hence, based on the Australian Rainfall and Runoff (Institution of Engineers Australia, 1987) a runoff coefficient of 6 is established for Murray Bridge for an average recurrence interval of 10 years.

Page 44 of 77 Riverbank Collapse Task 5 Report

The River Murray level data used in the analyses were obtained from Murray Darling Basin Authority for Murray Bridge at observation station (A4261162) incorporating data from No. 1 Pump Station (A4260522) and (A4261003) (Murray Darling Basin Authority, 2014). From these data, the river levels at Murray Bridge had remained relatively constant between 0.5 to 1 m AHD until the end of 2006, where they started to fall continuously and significantly to −1.5 m AHD until September 2009. They then gradually returned to the pre-2006 river levels. As shown in Figure 29, the study period from May 1, 2008 to February 28, 2009 (304 days) is selected during this period of low river levels to model the Long Island Marina riverbank collapse with the combined influence of river level fluctuation, rainfall and evaporation. For simplicity, the water level in the embayment area is assumed to be constant at 1 m AHD.

Figure 29: Daily river levels, daily rainfall and daily mean temperature from May 1, 2008 to February 28, 2009 at Long Island Marina.

As summarised in Figure 29, mean daily rainfall, as well as mean daily temperature, at the Long Island Marina site were collected from the Bureau of Meteorology (2014), in addition to evaporation and humidity, in order to provide climatic data for the riverbank stability modelling.

Within the SVFlux2D framework, the riverbank geometry is established from the DEM and ArcGIS, as mentioned above and shown in Figure 30. The boundary conditions are established to represent the actual site circumstances as: boundary AH was set to ‘climate’ which represents the combined effects of precipitation and evaporation over the 304 day study period; boundary HG is established as ‘head data review’ to account for the effects of river level variation over this period; and boundaries BH and

Page 45 of 77 Riverbank Collapse Task 5 Report

CF are set to the ‘no boundary’ condition, which indicates a barrier free condition for water flux; and the remainder are set to ‘zero flux’ to simulate no water flux through these boundaries.

(a) 2D cross-sectional model.

93 m 51 m

(b) 3D slope stability model. Figure 30: Results of (a) 2D (Day 282: February 6, 2008) and (b) 3D (Day 287: February 10, 2008) riverbank stability analyses of Long Island Marina site.

Page 46 of 77 Riverbank Collapse Task 5 Report

In the SVFlux3D model, in order to facilitate modelling pore water pressure variations due to river level fluctuation and climatic influences, the riverbank is divided into two zones, as shown in Figure 24(c). The climatic data of rainfall, evaporation, air temperature and humidity were assigned the ‘climate’ boundary condition to the ground surface of Zone I, while the river level fluctuation was applied as a ground surface boundary condition in Zone II as ‘head data review’; the sidewall boundary between Zones I and II was assigned the ‘no boundary condition’ to permit flux; while the remaining three sidewalls associated with both zones were assigned ‘zero flux’ to simulate no water flux.

4.4 Results The results of the 2D back analyses, using SVFlux2D and SVSlope2D, and those in 3D, using SVFlux3D and SVSlope3D, are presented as follow. The resulting pore water pressure and FoS time histories are examined with reference to river level fluctuation and climatic influences. As mentioned above, SVFlux is used to model the variation in pore water pressure (PWP) as a consequence of rainfall, evaporation and river level fluctuations based on the transient seepage model. As shown in Figure 30, 5 nodes were selected to monitor the PWP variation above the slip surface at locations T1, T2, T3, T4 and T5.

Figure 31 shows the evolution of PWPs as obtained from SVFlux2D at locations T1 – T5, with (solid lines) and without (dashed and dotted lines) the effects of evaporation. As indicated, the entire study period of 304 days is divided into two parts: Part A includes late autumn and winter (the rainy season in South Australia); and Part B the subsequent spring and summer seasons, which are warmer and drier.

Horizontally, the daily PWP profiles showed that although T1, T4 and T5 are at the same elevation, each has different pore water pressures due to the head variations between embayment and river level, but they exhibit a very similar trend throughout the 304-day period. This is also true when comparing T2 and T3. Specifically, at locations T1, T4 and T5, the profiles of PWP increased in late autumn and winter 2008 corresponding to rising river level, and then reduced in spring and summer 2008 as the river level dropped. The fluctuations in PWP profiles are caused by the climatic factors (i.e. rainfall, temperature and evaporation). The PWP at T2 and T3 remain unchanged during autumn and winter, but recorded increased in spring and summer, probably due to larger head variations between embayment and river level.

For T1, T4 and T5, there is very little effects of the evaporation on the evolution of PWPs evaporation in autumn and winter, but the effects are more pronounced in spring and summer, as the lines (with

Page 47 of 77 Riverbank Collapse Task 5 Report

Figure 31: Evolution of pore water pressure at 5 selected nodes through the entire research period accounting for, and without, evaporation. and without the influence of evaporation) have larger divergence at the beginning of spring (early September 2008). There is no effect of evaporation on T2 and T3 as soils at that depth are shielded from the elements. Therefore, the sensitivity to the combined effects of rainfall, evaporation and lateral bank seepage, as a consequence of river level fluctuation, diminishes with increasing depth. For example, as a consequence of the 3-day storm between December 12–14, 2008, there is immediate change in PWP at T1 than at T2, when comparing the PWPs before and after the storm.

During Part A, as indicated in Figure 31, roughly 164.7 mm rainfall was recorded (70% of that over the entire 304 day period), the river rose by approximately 0.3 m, and the lower daily temperatures resulted in modest daily evaporation, there is no to modest changes in PWP in each of the 5 nodes. In contrast, in Part B, a moderate rainfall (81.5 mm), accompanied by relatively high daily evaporations and a significant drawdown of river level (0.9 m), contributed to a significant reduction in pore water pressure. A summary of the PWP distribution is shown in Figure 32, with Figure 32(a) showing the PWPs during the day with the highest river level (Day 138: 15/09/2008) and Figure 32(b) illustrating the PWPs during the day with the lowest river level (Day 302: 26/02/2009).

As mentioned previously, deep-seated, circular slip failures in the soft and very soft clays of Holocene age were indicated as the dominant bank collapse mechanism along the lower River Murray. Therefore, back-analyses are performed using the limit equilibrium method (Bishop’s method of slices) implemented in SVSlope2D and SVSlope3D. Given that the riverbank collapse is recorded to have

Page 48 of 77 Riverbank Collapse Task 5 Report

(a)

(b) Figure 32: PWP distributions as a result of (a) the highest (Day 138) and (b) lowest (Day 302) river levels. occurred on February 4, 2009, the back analyses are benchmarked against this date, as well as the dimensions and extent of failure indicated by the bathymetric survey summarised in Figure 23(b). The geotechnical properties of the high plasticity, Silty Clay (CH) layer were varied marginally against the results of the site investigation (Table 5) until a FoS of 1.00 was obtained close to the recorded date of collapse.

Figure 33 presents the results of the 2D and 3D back analyses in terms of FoS as a function of date. The nearest boreholes and CPTus suggested an unsaturated Fredlund mode (Fredlund et al., 1996) for the Fill and CH layers, which is able to accommodate soil suction changes in the unsaturated bank materials, as shown by Equation (2).

Based on the models adopted and data obtained by SKM (2010b), the following models are incorporated in the analyses which follow: Fill layer: c' = 0 kPa; φ' = 28°; and CH layer: φ' = 0 kPa; φ' = 22°. The latter model aligns reasonably well with the single CU triaxial test performed by SKM (2010b)

Page 49 of 77 Riverbank Collapse Task 5 Report

(φ' = 27°), noting that only one soil sample was tested and the model adopted in the analyses utilises an average value of c' throughout the entire CH layer.

2.0

1.8 CRLM 2D Model 3D Model

1.6

1.4 FactorofSafety

1.2

1.0

0.8 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09

Date (dd/mm/yy)

Figure 33: Factors of safety from the 2D, 3D and CRLM models.

The results of the 2D analyses shown by the dashed line in Figure 33 are extremely encouraging. Firstly, the predicted date of failure is Day 282 (February 6, 2009) compares very favourably with the recorded date of failure February 4, 2009 (Day 280). Secondly, the FoS is consistently above 1.00 from the beginning of the analysis period until the predicted date of failure. Finally, the predicted slip surface is shown in Figure 30(a). The width of the collapsed region from the crest to the riverbank line is 19.8 m, which again compares extremely favourably with the actual width of 20 m determined from aerial photos [Figure 25(b)].

As mentioned above, a 3D slope stability analysis is performed to complement and enhance the 2D riverbank slope stability analyses. Generally, 2D riverbank slope stability analysis, which assumes that the failure surface is infinitely wide, is suitable for stability estimation or assessment over a relatively extensive region, such as adopting a large number of 2D cross-sections, spaced at appropriate intervals, to assess the stability of a large extent of the river. However, 3D riverbank slope stability

Page 50 of 77 Riverbank Collapse Task 5 Report analysis is generally more accurate for collapse back-analysis over relatively small regions where the dimensions or volume of the collapse zone are known (Stark and Eid, 1998). As shown in Figure 23(b), the bathymetric data clearly shows three, what appear to be, separate failures, identified as Regions (i) – (iii). From the DEWNR database of recorded riverbank collapse incidents, it is clear that Region (iii) has largest collapsed volume of soil, and hence attention is directed solely to this failure. The modelling of Regions (i) and (ii) is beyond the scope of this paper. The failure of Region (i), approximate a year prior to Region (iii), has some effect on the pre-collapse topography associated with Regions (ii) and (iii) and, given the limited extent of available observational and survey data, presents significant challenges and uncertainties, therefore, the effects are not quantified and ignored in this analysis. Figure 33 also shows, by means of a solid line, the results of the 3D back- analysis of the riverbank. As for the 3D analysis, an unsaturated Fredlund model is again adopted with the following properties: c' = 0 kPa; φ' = 20°. As can be seen, this 3D model omits the Fill layer due to insufficient information and large computation resources required (additional finite elements) to incorporate this thin layer of Fill into the model. The omission represents a modest reduction in soil weight and cohesion when compared with the 2D model and also the reduction in φ' from 24° to 22° in the 3D model.

As shown in Figure 33, the reductions in FoS can be observed in the 3D model, which likely relate to the reduction in the daily river level and the effects of large rainfall events. As can be observed, river level fluctuation dominates the likelihood of FoS variation in both 2D and 3D. Compared with the 2D model, the 3D model FoS appears to be less sensitive to river level variation, however a significant drawdown may result in a significant reduction in FoS. Again, the results represent extremely good correlation between the predictions and observations. Firstly, the predicted date of failure Day 291 (February 15, 2009) aligns well with the recorded date of failure February 4, 2009 (Day 280). It is noted that the extra loads applied by the three vehicles that slipped into the river are not included in the model. The FoS is predicted to be close to unity on February 4, 2009, and coincidentally, the additional vehicle weights might well be sufficient for failure to occur on that day. Secondly, the FoS is again consistently above the unity from the beginning of the analysis period until close to the predicted date of failure. Finally, the predicted volume of the slip surface is in Figure 9(b) and the volume and surface area of the collapse region is comparable to the results of the bathymetric survey.

In order to understand better the relationship between rainfall, river level fluctuations and riverbank stability, additional 2D analyses are performed using SVFlux and SVSlope. The first series of analyses examines the situation where the river level remains static while rainfall continues to vary in accordance with the historical record. The second series examines the situation of extreme rainfall events. Figure 33 and Figure 34 present the results of analyses adopting a constant river level model (CRLM) with a static level at –0.521 m AHD for the entire 304 day study period. This level coincides with the initial river level associated with the actual historical model (HM) examined in the previous

Page 51 of 77 Riverbank Collapse Task 5 Report analysis. As can be seen from Figure 33, compared with the HM 2D and 3D factors of safety, the CRLM FoS generally remains constant and consistently greater than the unity, with several minor reductions as a result of intensive rainfall events.

Figure 34(a) presents the time histories of FoS and PWP in both the HM and CRLM in mid-December 2008, when the most intense storm of the 304 study day period resulted in 29.8 mm of rainfall in 3 days. It can be observed that the PWPs associated with the CRLM at location T1 (refer to Figure 31(a) are considerably lower than those derived from the HM. This is because the lower river level (–0.521 m AHD) creates a depressed groundwater table resulting in decreased PWPs. It can also be observed in Figure 34(a), that the riverbank is less stable in the HM than the CRLM on 11 December due to the lower river level, but more stable in the HM than in the CRLM after the storm. The river levels during these 5 days (10–14 December) were –0.439, –0.486, –0.421, –0.278 and –0.344 m AHD, respectively. It can also be seen clearly that the CLRM FoS is generally constant throughout mid-December 2008, albeit with a negligible reduction after the storm. This demonstrates that daily river level fluctuation has an observable influence on the FoS, while the effect of daily rainfall is modest. In contrast, relatively minor precipitation was recorded (0.8 mm) during the intervening period leading up to the collapse on 4 February 2009. As shown in Figure 13(b), the HM FoS continues to decline until collapse is predicted on 6 February 2009. During this period the CRLM FoS remains constant and consistently greater than the unity.

Further examining the influence of rainfall on riverbank collapse, an extreme rainfall scenario is modelled. The most significant storm on record at Murray Bridge occurred in January 1941 when 189.6 mm of rain fell over a period of 5 days. This is more than 6 times greater than the storm examined above.

(a) HM under the storm occurred in Dec. 2008. (b) MRM with high daily river levels (approx. –0.2 m AHD).

Page 52 of 77 Riverbank Collapse Task 5 Report

(c) MRM with medium daily river levels (approx. (d) MRM with low daily river levels (approx. –0.5 m AHD). –0.8 m AHD). presents the results of the magnified rainfall model (MRM), which incorporates the January 1941 storm. Four different scenarios are presented: Figure 35(a) shows the results of the HM for benchmarking purposes; whereas Figure 35(b) – (d) present the results of the MRM with high (–0.2 m AHD in September 2008), medium (–0.5 m AHD in May 2008) and low (–0.8 m AHD January 2009) river levels, respectively.

Figure 35(a) shows the behaviour of the HM as a result of the historical storm that occurred in December 2008. The FoS remains consistently above unity with daily river levels fluctuating around –0.5 m AHD. In contrast, the FoS resulting from the MRM with high and medium daily river levels, as shown in Figure 35(b) and (c) respectively, reduces significantly with collapse predicted. The extreme storm results in an approximate increase of 30 kPa in PWP, and consequently triggers a rapid and significant decrease in FoS in both scenarios.

Page 53 of 77 Riverbank Collapse Task 5 Report

(a) Storm in mid-Dec. 2008.

(b) Scenario of the riverbank collapse at Long Island Marina on 4 Feb. 2009. Figure 34: Factors of safety for historical model (HM) and constant river stage model (CRLM) in two scenarios.

Page 54 of 77 Riverbank Collapse Task 5 Report

(a) HM under the storm occurred in Dec. 2008. (b) MRM with high daily river levels (approx. –0.2 m AHD).

(c) MRM with medium daily river levels (approx. (d) MRM with low daily river levels (approx. –0.5 m AHD). –0.8 m AHD). Figure 35: Magnified rainfall model (MRM) under different river level scenarios.

Figure 35(d) shows a similar outcome to that of the HM in Figure 35(a), where the FoS remains relatively consistent above unity, in the former as a result of the MRM with low daily river levels. After the extreme rainfall event, the change in PWP at T1 is 20 kPa, which is roughly 10 kPa less than that observed in the high and medium river level scenarios in Figure 35(b) and (c).

A sudden or rapid drawdown scenario has also been examined based on the HM. The river level was modelled to decrease from the highest point (–0.005 m AHD, Day 138 of HM) to the lowest (–1.002 m AHD, Day 302 of HM) within 2 days (drawdown ratio = 0.5 m/day) and 5 days (drawdown ratio = 0.2 m/day). The results show that, the greatest drawdown ratio (0.5 m/day) is associated with the lowest FoS (0.88), followed by drawdown ratio = 0.2 m/day (FoS = 0.914). The HM (164-day drawdown period with a drawdown ratio = 0.006 m/day) has the highest relative FoS of these three scenarios, albeit less than unity.

The analyses presented in this section have demonstrated that the river level fluctuations have a far greater influence on riverbank collapse than rainfall. However, if an extreme rainfall event coincides

Page 55 of 77 Riverbank Collapse Task 5 Report with a medium (–0.5 m AHD) to high (–0.2 m AHD) river level, collapse is likely to occur. Whilst not a key objective of the present study, a series of additional analyses have also been performed to examine the influence of evaporation on FoS. The results, however, demonstrate that evaporation has a marginal effect on FoS.

4.5 Summary In summary, this section has sought to: (i) model the riverbank collapse incident, in both 2D and 3D, which occurred at Long Island Marina, Murray Bridge, South Australia, on February 4, 2009; (ii) examine the influence and sensitivity of river level fluctuations and climatic factors on riverbank stability; and (iii) determine the dominant triggers affecting collapse. The modelling has been undertaken using the limit equilibrium method in 2D and 3D using SVFlux and SVSlope. To facilitate the stability analyses, a GIS framework has been adopted, which involved: examination of the recorded collapsed regions by visual interpretation of high-resolution aerial images; obtaining topographic information of the site from a digital elevation model; and calculation of the dimensions of the collapsed regions from high-resolution aerial images. Back-analyses have been performed using geotechnical data obtained from a site investigation incorporating both in situ and laboratory testing and validated against data from the recorded collapse.

This study has shown that: (a) Both the adopted process of 2D and 3D stability modelling yielded excellent predictions of the collapse when compared against the recorded date of collapse and dimensions of the failed region. The adopted geotechnical model parameters align with those derived from the site investigation. (b) The integration of GIS with transient unsaturated soil modelling has proved to be an effective tool for accurately predicting riverbank stability. (c) River fluctuations, rather than climatic factors, dominate the likelihood of riverbank collapse along the Lower River Murray. (d) Sudden or rapid drawdown is associated with a lower FoS compared with the historical model. (e) However, extreme rainfall events, coinciding with medium to high river levels, are also likely to trigger riverbank collapse.

Page 56 of 77 Riverbank Collapse Task 5 Report

5 Identifying the Areas Susceptible to Riverbank Instability

5.1 Introduction In the Task 4 report entitled Mechanisms and Processes of the Millennium Drought Riverbank Failures (Hubble and De Carli, 2015) it was stated that the majority of the larger bank failure features are associated with deep holes that have been eroded into the channel floor. These deep holes are formed due to either: (a) bedrock constriction and pronounced narrowing of the channel cross- section; or (b) large outcrops of bedrock which protrude up from the floor of the channel (Hubble and De Carli, 2015). This has generated erosive flow patterns during periods of higher flow that have eroded or scoured deep holes in the channel and over-steepened the channel margins and banks. Riverbanks in reaches of the channel affected in this way are less stable than the banks present elsewhere along the Lower Murray where the channel profile is wider and shallower (Hubble and De Carli, 2015).

Hubble and De Carli (2015) also identified the failure prone sites which possess the geomorphologic features of deep holes through bathymetric studies, namely: Thiele Reserve, Bells Reserve, Whitesands, Sturt Reserve, Long Island Marina, Woodlane Reserve and East Front Road near Younghusband. Over-steepened riverbanks and local anthropogenic modification of the banks (e.g. placement of fill or construction of the embankments near the waterline), combined with low pool- levels, are identified as the main causes or factors affecting riverbank stability. Numerical modelling was undertaken by Hubble and De Carli (2015) to study the effects of the aforementioned factors on the FoS of the riverbanks at the Thiele Reserve.

This section of the report will expand the numerical studies to determine the FoS of the riverbanks at Thiele Reserve when the river level is at normal pool level, i.e. 0 m AHD. Geographic information system (GIS) modelling and mapping has again been used to facilitate the study. A methodology, combining both SVSlope and ArcGIS, has been developed to identify the riverbanks along the Lower River Murray that are susceptible to instability.

5.2 Methodology and Results As demonstrated in the earlier sections slope failure susceptibility prediction, at medium or large scales, GIS modelling and mapping greatly enhances and simplifies the analyses. ArcGIS can assist in locating the deeps holes and over-steepened riverbanks, an example of which is shown in Figure 36. By extracting the contour lines from the DEM using ArcGIS [Figure 36(b)], and by using its filtering features, the contour lines at or below –10 m AHD [Figure 36(c)] can be highlighted to identify the location of the deep holes near Thiele Reserve. The same process can be applied to the Lower River Murray and the results are shown in Figures Figure 36 to 39.

Page 57 of 77 Riverbank Collapse Task 5 Report

N (a)

N

(b)

Avoca Dell

N

Thiele Reserve

Murray Bridge East

(c) Murray Bridge

Figure 36: Locating deep holes near Thiele Reserve using: (a) bathymetry from Hubble and De Carli (2015) (areas of bedrock margins are represented by dashed white line); (b) a contour map obtained from ArcGIS superimposed on satellite imagery from Google Maps; (c) blue contour lines showing elevations that are –10 m AHD or deeper, whilst the green contour lines represent 0 m AHD.

Page 58 of 77 Riverbank Collapse Task 5 Report Lock 1

Blanchetown

Blanchetown

Swan Reach

Punyelroo Big Bend Wongulla Big Bend

Nildottie

Wongulla Walker Flat

Walker Flat

Figure 37: Locations of deepened channel beds downstream from Blanchetown to Walker Flat (blue contour lines show elevations that are at –10 m AHD or deeper, whilst the green contour lines are at 0 m AHD).

Page 59 of 77 Riverbank Collapse Task 5 Report

Younghusband Walker Flat Caloote Wall Flat

Caurnamont Purnong Woodlane

Younghusband

Bowhill

Mannum

Mypolonga

Caloote Caurnamont

Wall Flat Purnong

Woodlane

Mypolonga Avoca Dell

Murray Bridge East Avoca Dell Bowhill Murray Bridge East

Figure 38: Locations of deepened channel beds downstream from Walker Flat to Murray Bridge East (blue contour lines show elevations that are at –10 m AHD or deeper, whilst the green contour lines are at 0 m AHD).

Page 60 of 77 Riverbank Collapse Task 5 Report

Monteith

Thiele Reserve Murray Bridge East Murray Bridge

White Sands

Thiele Reserve

Murray Bridge East Monteith Tailem Bend

White Sands

Jervois Sturt Reserve Woods Point Long Flat

Long Island Marina Murray Bridge Tailem Woods Bend Point Jervois

Wellington East Wellington Wellington East Wellington

Figure 39: Locations of deepened channel beds downstream from Murray Bridge East to Wellington (blue contour lines show elevations that are at –10 m AHD or deeper, whilst the green contour lines are at 0 m AHD).

Page 61 of 77 Riverbank Collapse Task 5 Report

The results show that there are a number of deep holes from Blanchetown to Purnong and they are generally separated by large distances (refer to Figures 37 and Figure 38). Deepened channels are more common from Bowhill to Younghusband, and from Mannum to Wellington (see Figure 38 and 39). As a case study, the effects of the deep holes on the FoS of the riverbanks near Thiele Reserve will be examined, along with other contributing factors.

By using ArcGIS, a series of transects has been created along the waterline, at the normal pool level of 0 m AHD near Thiele Reserve (Figure 40), and the elevations along the transects have been extracted using the bilinear interpolation method at a one-metre interval. This approach yields accurate cross- sections, suitable for subsequent importing into SVSlope2D.

Thiele Reserve

Murray Bridge East

Figure 40: A series of transects along 0 m AHD contour lines near Thiele Reserve (blue contour lines show elevations that are at –10 m AHD or deeper, whilst the green contour lines are at 0 m AHD. The red dot marks the approximate location of the past riverbank instability incident and the dashed orange line marks the areas of the bedrock margins).

As in the previous sections, SVSlope2D, which performs limit equilibrium slope stability calculations in 2D, has been adopted to study riverbank stability. The geotechnical profile and the parameters are determined from a combination of field and laboratory testing. A single onshore CPTu has been undertaken (Jaksa et al., 2015), and the results showed that the subsurface profile encountered at

Page 62 of 77 Riverbank Collapse Task 5 Report

Thiele Reserve is generally comprised of Fill overlying Silty CLAY with a transition occurs at around 0.8 m AHD. The layer of dark-grey, very-soft Silty CLAY with high plasticity fines, is relatively thick, extending down to –14.3 m AHD (Jaksa et al., 2015). The investigation confirmed the expected Quaternary aged Alluvial Flat Deposits, as seen on the Geological Survey of South Australia (1962) Barker map Sheet 1 54-13. A Clayey SAND and Sandy CLAY interbedded layer underlies the Silty CLAY, extending to the refusal depth. Furthermore a series of consolidated undrained (CU) triaxial tests were carried out on soil samples collected from Thiele Reserve using the procedures recommended by Head (1998). Drained parameters c' and ' were found to be 0 kPa and 41 respectively (Jaksa et al., 2015). A summary of the geotechnical parameters adopted for numerical modelling is presented in Table 7.

Table 7: Adopted geotechnical parameters for stability calculations.

Layer Elevation (m AHD) c' (kPa) φ' (°) γ (kN/m3)

Fill Ground surface to −0.8 0 28° 18

Silty Clay (CH) −0.8 to −14.3 0 41° 16

Sand > −14.3 0 30° 17

Subsequently, the bank profiles at Thiele Reserve are modelled as three separate, homogenous soil layers. For simplicity, the soil layers and the transitions zone between the soil layers are assumed to be horizontal. An effective stress analysis was performed in the Fill, CH and Sands layer using the Mohr-Coulomb failure criterion. Effective stress analysis was used to assess the long-term stability of the riverbanks. An example of the soil profile, as well as geotechnical parameters used in the analysis is shown in Figure 41 and an example of the result is shown in Figure 42.

Stability analyses have been undertaken from Transects No. 4207 to 4242, and also No. 5527 to 5541 using the soil profiles (Fill-Clay-Sand) outlined previously. The results of the analysis are summarised in Figure 43. Transects No. 5510 to 5526, where the bedrock margin is located (represented by dashed orange line in Figure 40), have not been analysed. This is because the subsurface of these transects has different geotechnical profiles than that associated with hard bedrock and slip circle analysis is unsuitable for this type of subsurface material. The yellows dots shown in Figure 43 represent the locations of the riverbanks that are vulnerable to instability in the long-term and also the locations of the top of slipe circles, normally at the crest of the riverbanks and very close to the 0 m AHD waterline. The locations of the levees are also shown as thick green continuous lines in Figure 43. It is noted that, in contrast to the earlier work, the climatic influences, such as rainfall and evapotranspiration, are ignored in this study due to the assumption of steady state conditions and to simplify the analyses. No external loads (e.g. buildings, road traffic, etc.) have been considered.

Page 63 of 77 Riverbank Collapse Task 5 Report

Figure 41: An example of an SVSlope2D analysis (Transect No. 4212).

Figure 42: An example of the result of an SVSlope2D analysis (transect no. 5529). Page 64 of 77 Riverbank Collapse Task 5 Report

Figure 43: Riverbanks that are vulnerable to slumping are identified (shown in yellow dots) after the analyses (assuming a river level at 0 m AHD; blue contour lines show elevations that are at –10 m AHD or deeper with one metre interval; red dot shows the approximate location of the occurrence of past slumping; dashed yellow lines are areas of bedrock margins; and levees are represented by the thick green line).

In this case study, riverbanks with FoS below 1.05 are considered to be vulnerable and could collapse in the long-term. Each result of the numerical modelling was studied in detail, and it is concluded that shallow/planar type failures, not massive rotational or deep seat failures, can be expected to occur in the long-term when the pool level is held constant at 0 m AHD. The primary cause of these potential failures is the over-steepened riverbanks, which have slope angles much larger than the ' of the Silty CLAY (i.e. 41°). The riverbanks with high slope angles, which appear to be stable in the short-term due to the influence of total suction, may collapse in extreme rainfall events in the longer term. The over-steepened riverbank can be found, in this case and predominantly near the deep holes and also on the outside bend of meanders, are vulnerable to accelerated erosion. The placement of the Fill layer, which acts as surcharge, is found to be the secondary factor, whilst the levees or embankments, which are located >10 metres away from the waterline, are very unlikely to contribute to potential failure.

Page 65 of 77 Riverbank Collapse Task 5 Report

Therefore, characterising the topographies of the riverbanks, such as the inclination angle of the riverbank and the thickness of fill above pool level, using ArcGIS, can be used as indicators to aid with high-risk areas identification. In the evaluation of landslide and riverbank slope failure in wide natural slopes, it is typical to assume the slope surface is perfectly level. However, natural slopes are generally undulating and irregular, an example is shown in Figure 42 and, as a consequence, the value of the inclination can vary depending on the size of the grid that ArcGIS used to determined it. In order to illustrate the effect of grid size, the inclination of the riverbanks near Thiele Reserve have been mapped using the DEMs within ArcGIS and presented in Figures 44 and 45. Two different sizes (or matrix) of grids: 1 × 1 in Figure 44 and 10 × 10 in Figure 45, are used in the DEM to evaluate the inclination of the riverbank slope. By comparing Figures 44 and 45, it is observed that the size of the adopted grid can greatly influence the evaluation of the inclination. For example, it can be seen that flatter (10° to 30°) inclinations are obtained for vulnerable riverbanks when a 10 × 10 matrix of grids is used (Figure 44), while a more detailed and variable (from a low value of 10° to more than 45°) inclination map is obtained with a finer grid (e.g. 1 × 1 matrix of grids as shown Figure 44). Figure 44 clearly shows that the high-risk riverbanks are associated with high inclination values; therefore, the inclination map determined by using a finer grid is accurate and more suitable to be used in the identification of high-risk areas.

5.3 Summary In summary, this section has shown that: (i) deep holes can be determined using DEM with ArcGIS; (ii) deep holes can be found from Blanchetown to Wellington, but there a greater number of deep holes, and they are longer and more widespread, downstream of Mannum; and (iii) the riverbanks near Thiele Reserve, with slope angles greater than 40° are vulnerable and could collapse in the long-term, but they are most likely to be shallow or planar failure types. The modelling has been undertaken using the limit equilibrium method in SVSlope2D. To facilitate the stability analyses, a GIS framework has been adopted and topographic information of the site has been obtained from a digital elevation model (DEM). Finally, topographical factors, such as the inclination of the slope, can be used as a very effective indicator for identifying riverbank instability near Thiele Reserve. The Task 6 report will undertake further study and expand the study to include other locations.

Page 66 of 77 Riverbank Collapse Task 5 Report

45°

0°

Figure 44: Inclination of the riverbanks near Thiele Reserve using a 1 × 1 matrix of grids.

Page 67 of 77 Riverbank Collapse Task 5 Report

45°

0°

Figure 45: Inclination of the riverbanks near Thiele Reserve using a 10 × 10 matrix of grids.

Page 68 of 77 Riverbank Collapse Task 5 Report

6 Summary

This report presents the methodologies that have been developed and used by the authors to back- analyse a number of historical collapses, was well as tools developed to assess the stability of the riverbanks stabilities and identify the high-risk areas along the Lower River Murray. The geographic information system (GIS) technique, implemented using ArcGIS, has greatly facilitated riverbank instability research by providing a number of useful functions for spatial data management. The GIS- based technique simplifies the process of identifying collapsed regions (locations, size and volume), deepened riverbeds, over-steepened channels and the geometry and cross-sections of the riverbanks of the Lower River Murray. Satellite images, aerial photographs and digital elevation models (DEMs) have been provided by the South Australian Department of Environment, Water and Natural Resources (DEWNR) for this research, and these high-resolution, remotely sensed data have been processed using ArcGIS and studied extensively throughout this research.

This report has demonstrated that by integrating GIS-based methods and slope analysis tool (both ArcGIS and SVSlope), riverbank collapses can be accurately and reliably predicted by using the proposed frameworks. The effects of the rainfall and groundwater flow on the stability of the riverbank has also been studied by integrating a groundwater and unsaturated soil modelling tool, called SVFlux with the slope stability analysis software, SVSlope, both in 2D and 3D. The changes in groundwater level and flow correspond to the changes of the river pool level. The moisture content of the soil, which is affected by rainfall, evapotranspiration and ground water flow/level influences the soils’ suction and shear strength, and ultimately riverbank stability. By validating the models, the studies have demonstrated that the proposed framework provides reliable riverbank stability predictions. The integration of GIS with high-resolution spatial data facilitates the process of collapsed region identification, model geometry development and the calculation of the dimensions of the collapsed regions.

The accuracy of the stability prediction hinges on the accuracy of soils parameters. Many factors can influence the accuracy of the soil parameters determined from field and laboratory testing, such as the field sampling technique, natural variability of the soil properties, the number of samples, sample preparation and quality control in the laboratory. Therefore, back-analyses have been undertaken at 4 collapsed sites. The results of the back-analysed soil shear strengths at the 4 sites have shown great consistency with those obtained from site investigations undertaken at locations adjacent to the collapse sites.

Both 2D and 3D stability modelling yielded excellent predictions of the collapse when compared against the recorded dates of collapse and dimensions of the failed region. However, due to project time constraints, 2D analyses will be adopted for the Task 6 report, as the 2D modelling only requires modest resources and provides reliable yet conservative estimates of FoS.

Page 69 of 77 Riverbank Collapse Task 5 Report

7 References

Abramson, LW, Lee, TS, Sharma, S and Boyce, GM, 2002, ‘Slope stability and stabilization methods’, 2nd edition, John Wiley & Sons, Inc., New York. Acharya, G, De Smedt, F, and Long, N, 2006, ‘Assessing landslide hazard in GIS: a case study from Rasuwa, Nepal’. Bulletin of Engineering Geology and the Environment, 65(1), 99–107. Arnott, C, 1994, ‘Rates of stage fluctuation on the Murray River between Hume reservoir and Albury’. Honours research thesis, University of Melbourne. Aronoff, S, 1989, ‘Geographic Information System: A Management Perspective’. WDL Publications, Ottawa, 279 pp. Beal, A, Brown, R and Toole, JO, 2010, ‘Expert panel workshop – Riverbank collapse hazard program’. (unpublished). Berilgen, MM, 2007, ‘Investigation of stability of slopes under drawdown conditions’. Computers and Geotechnics, 34(2), 81–91. Bouwer, H, 1978, ‘Groundwater hydrology’. McGraw-Hill, New York. Bozzano, F, Martino, S, Montagna, A and Prestininzi, A, 2012, ‘Back analysis of a rock landslide to infer rheological parameters’. Engineering Geology, 131-132, 45–56. Brand, EW, 1984, ‘Landslides in Southeast Asia: A state-of-the-art report’. Proceedings of 4th Int. Symp. on Landslides, Toronto, 17–59. Budhu, M and Gobin, R, 1995, ‘Seepage-induced slope failure on sandbars in grand canyon’. Journal of Geotechnical Engineering, 121(8), 601–609. Bureau of Meteorology, 2014, ‘Climate data online’, Commonwealth of Australia, . Burns, SE and Mayne, PW, 1998, ‘Monotonic and dilatory pore pressure decay during piezocone tests in clay’. Canadian Geotechnical Journal, 35, 1063–1073. Burrough, PA, 1986, ‘Principles of geographical information systems for land resources assessment.’ Clarendon Press, Oxford, 194 pp. Cai, MF, Xie, MW and Li, CL, 2007, ‘GIS-based 3D limit equilibrium analysis for design optimization of a 600 m high slope in an open pit mine’. Journal of University of Science and Technology Beijing, 14(1), 1–5. Calabresi, G, Colleselli, F, Danese, D, Giani, GP, Mancuso, C and Montrasio, L, 2013, ‘A research study of the hydraulic behaviour of the Po river embankments’. Canadian Geotechnical Journal, 50(9), 947–960. Carrara, A, Cardinali, M, Detti, R, Guzzetti, F, Pasqui, V and Reichenbach P, 1991, ‘GIS techniques and statistical models in evaluating landslide hazard’. Earth Surface Processes and Landforms, 16, 427–445.

Page 70 of 77 Riverbank Collapse Task 5 Report

Carrara, A, Cardinali, M, Guzzetti, F and Reichenbach, P, 1995, ‘GIS-based techniques for mapping landslide hazard.’ In GIS in assessing natural hazards, Eds. A Carrara, A and F Guzzetti, Kluwer Publications, Dordrecht, 135–176. Casagli, N, Rinaldi, M, Gargini, A and Curini A, 1999, ‘Pore water pressure and streambank stability: Results from a monitoring site on the Sieve River, Italy’. Earth Surface Processes and Landforms, 24(12), 1095–1114. Coffey 2012, ‘Review of management options for four riverbank collapse high risk sites’. 07093AA-AD Final Report, December 12, Adelaide. Dahm, J and Edwards M, 1988, ‘Preliminary investigation of the influence of Matahina Power Station on river bank stability along the Rangitaiki River’. Electricity Corporation of New Zealand Ltd. Hamilton. DFW, 2010. ‘Riverbank collapse hazard expert panel, submission: site reports & investigation’. Department for Water, Government of South Australia. (unpublished). DFW, 2011, ‘Lower River Murray riverbank collapse hazard site status report May 2011’. Filz, GM, Brandon, TL, and Duncan, JM, 1992, ‘Back analysis of the Olmsted Landslide using anisotropic strengths’. Transportation Research, Record 1343, TRB, National Research Council, 72–78. Fredlund, DG, 2006, ‘Unsaturated soil mechanics in engineering practice’. Journal of Geotechnical & Geoenvironmental Engineering, 132(3), 286–321. Fredlund, DG and Xing, A. 1994, ‘Equations for the soil-water characteristic curve’. Canadian Geotechnical Journal, 31(4), 521–532. Fredlund, DG, Xing, A and Huang, S, 1994, ‘Predicting the permeability function for unsaturated soils using the soil-water characteristic curve’. Canadian Geotechnical Journal, 31(4), 533–546. Fredlund, DG, Xing, A, Fredlund, MD and Barbour, SL, 1996, ‘The relationship of the unsaturated soil shear to the soil-water characteristic curve’. Canadian Geotechnical Journal, 33(3), 440–448. Fredlund, DG, Xing, A and Huang, S, 1995, ‘Predicting the permeability function for unsaturated soils using the soil-water characteristic curve’. Canadian Geotechnical Journal, 31, 533–546. Gilbert, RB, Wright, SG and Liedtke, E, 1998, ‘Uncertainty in back analysis of slopes: Kettleman Hills case history’. Journal of Geotechnical & Geoenvironmental Engineering, 124(12), 1167–1176. Green, SJ, 1999, ‘Drawdown and river bank stability’. Masters research thesis, Dept. of Civil and Environmental Engineering, University of Melbourne. Harris, SJ, Orense, RP and Itoh, K, 2012, ‘Back analyses of rainfall-induced slope failure in Northland Allochthon formation’. Landslides, 9(3), 349–356. Hashimoto, A, Oguchi, T, Hayakawa, Y, Lin, Z, Saito, K and Wasklewicz, TA, 2008, ‘GIS analysis of depositional slope change at alluvial-fan toes in Japan and the American Southwest’. Geomorphology, 100(1-2), 120–130. Higgins, DT, 1980, ‘Unsteady drawdown in 2D water-table aquifer’. Journal of the Irrigation and Drainage Division, ASCE, 106(3), 237–251.

Page 71 of 77 Riverbank Collapse Task 5 Report

Hong, Y, Adler, R and Huffman, G, 2007a, ‘Use of satellite remote sensing data in the mapping of global landslide susceptibility’. Natural Hazards, 43(2), 245–256. Hooke, JM, 1979, ‘An analysis of the processes of river bank erosion’. Journal of Hydrology, 42(1-2), 39–62. Hubble, T and De Carli, E, 2015, ‘Mechanisms and Processes of the Millennium Drought River Bank Failures: Lower Murray River, South Australia’, Goyder Institute for Water Research Technical Report Series No. 15/5, Adelaide, South Australia. Institution of Engineers, Australia, 1987, ‘Australian rainfall and runoff: A guide to flood estimation’. Vol. 1, Editor-in-chief D.H. Pilgrim, Revised Edition 1987 (Reprinted edition 1998), Barton, ACT. Jaksa, MB, Kuo, YL, Hubble, T, De Carli, E and Liang, C, 2015, ‘Geotechnical investigations and data acquisition’, Goyder Institute for Water Research Technical Report Series No. 15/x, Adelaide, South Australia. (In press.) Jia, GW, Zhan, TLT, Chen, YM and Fredlund, DG, 2009, ‘Performance of a large-scale slope model subjected to rising and lowering water levels’. Engineering Geology, 106(1-2), 92–103. Kamp, U, Growley, BJ, Khattak, GA and Owen, LA, 2008, ‘GIS-based landslide susceptibility mapping for the 2005 Kashmir earthquake region’. Geomorphology, 101(4), 631–642. Lane, PA, and Griffiths, DV, 2000, ‘Assessment of stability of slopes under drawdown conditions’. Journal of Geotechnical & Geoenvironmental Engineering, 126(5), 443–450. Lee, S, Choi, J and Min, K, 2004, ‘Probabilistic landslide hazard mapping using GIS and remote sensing data at Boun, Korea’. International Journal of Remote Sensing, 25(11), 2037–2052. Li, XP, Tang, HM and Chen, S, 2008, ‘A GIS-supported logistic regression model applied in regional slope stability evaluation’. Landslides and Engineered Slopes: From the Past to the Future, vol. 1 and 2, 789–794. Li, YQ, Feng, Y and Jiang, SH, 2011, ‘Stability analysis of soil slope during rapid drawdown of water level’. Proceedings of 2nd Int. Conf. on Mechanics Automation and Control Engineering (MACE), July 15-17, Hohhot, China, 3454–3457. Liang C, Jaksa, MB and Ostendorf, B, 2012, ‘GIS-based back analysis of riverbank instability in the lower River Murray’. Australian Geomechanics, 47(4), 59–65. Liang, C, Jaksa, MB, Kuo, YL and Ostendorf, B, 2015a, ‘Back analysis of Lower River Murray riverbank collapses’. Australian Geomechanics, 50(2), 29–41.

Liang, C, Jaksa, MB, Ostendorf, B and Kuo, YL, 2015b, ‘Influence of river level fluctuations and climate on riverbank stability’. Computers and Geotechnics, 63(1), 83-98. Liang, C, Jaksa, MB, Kuo, YL and Ostendorf, B, 2015c, ‘Identifying areas susceptible to high risk of riverbank collapse along the Lower River Murray’. Computers and Geotechnics, 69(9), 236–246.

Lumb, P, 1975, ‘Slope failures in Hong Kong’. Quarterly Journal of Engineering Geology and Hydrogeology, 8(1), 31–65.

Page 72 of 77 Riverbank Collapse Task 5 Report

Magliulo, P, Di Lisio, A, Russo, F and Zelano, A, 2008, ‘Geomorphology and landslide susceptibility assessment using GIS and bivariate statistics: a case study in southern Italy’. Natural Hazards, 47(3), 411–435. Marble, DF, 1990, ‘Geographic information systems: an overview – Introductory readings in geographic information systems’. Eds. D. F. Marble and D. J. Peuquet, Taylor and Francis Ltd., New York, 9–17. Mayo, D, 1982, ‘Permissible rates of river rise and fall below Dartmouth Regulating Dam.’ State Electricity Commission of Victoria. Merwade, VM, 2007, ‘An automated GIS procedure for delineating river and lake boundaries’. Transactions in GIS, 11(2), 213–231. Mukhlisin, M and Taha, M, 2012, ‘Numerical model of antecedent rainfall effect on slope stability at a hillslope of weathered granitic soil formation’. Journal of the Geological Society of India, 79(5), 525–531. Murray Darling Basin Authority, 2014, ‘Live river data – Murray river border to the sea’, . National Water Commission, 2005, ‘Australian water resources 2005: A baseline assessment of water resources for the National Water Initiative’, 115 pp., . Okagbue, C and Abam, T. 1985. ‘An analysis of stratigraphic control on the river bank failure.’ Engineering Geology, 22(3), 231–245. Okui, Y, Tokunaga, A, Shinji, M and Mori, S, 1997, ‘New back analysis method of slope stability by using field measurements’. Int. Journal of Rock Mechanics and Mining Sciences, 34(3–4), 234–249. Osman, AM and Thorne, CR, 1988, ‘Riverbank stability analysis 1: Theory’. Journal of Hydraulic Engineering, ASCE, 114(2), 134–150. Ozdemir, A, 2009, ‘Landslide susceptibility mapping of vicinity of Yaka Landslide (Gelendost, Turkey) using conditional probability approach in GIS’. Environmental Geology, 57(7), 1675–1686. Pantha, BR, Yatabe, R and Bhandary, NP, 2008, ‘GIS-based landslide susceptibility zonation for roadside slope repair and maintenance in the Himalayan region’. Episodes, 31(4), 384–391. Pauls, GJ, Sauer, EK, Christiansen, EA and Widger, RA, 1999, ‘A transient analysis of slope stability following drawdown after flooding of a highly plastic clay’. Canadian Geotechnical Journal, 36(6), 1151–1171. Rahardjo, H, Leong, EC and Rezaur, RB, 2008, ‘Effect of antecedent rainfall on pore-water pressure distribution characteristics in residual soil slopes under tropical rainfall’. Hydrological Processes, 22(4), 506–523. Rahardjo, H, Li, XW, Toll, DG and Leong, EC, 2001, ‘The effect of antecedent rainfall on slope stability’. Geotechnical & Geological Engineering, 19(3-4), 371–399.

Page 73 of 77 Riverbank Collapse Task 5 Report

Rahardjo, H, Ong, TH, Rezaur, RB and Leong, EC, 2007, ‘Factors controlling instability of homogeneous soil slopes under rainfall’. Journal of Geotechnical & Geoenvironmental Engineering, 133(12), 1532–1543. Rahimi, A, Rahardjo, H and Leong, EC, 2011, ‘Effect of antecedent rainfall patterns on rainfall-induced slope failure’. Journal of Geotechnical and Geoenvironmental Engineering, 137(5), 483–491. Ray, RL and Smedt, FD 2009, ‘Slope stability analysis on a regional scale using GIS: A case study from Dhading, Nepal’. Environmental Geology, 57(7), 1603–1611. Rinaldi, M and Casagli, N, 1999, ‘Stability of streambanks formed in partially saturated soils and effects of negative pore water pressures: The Sieve River (Italy)’. Geomorphology, 26(4), 253–277. Rinaldi, M, Casagli, N, Dapporto, S and Gargini, A, 2004, ‘Monitoring and modelling of pore water pressure changes and riverbank stability during flow events’. Earth Surface Processes and Landforms, 29(2), 237–254. Schuster, RL and Krizek, RJ, 1978, ‘Landslides: Analysis and control’. Transportation Research Board, National Research Council, Special Report No. 176. SKM 2010a, ‘Study into river bank collapsing for Lower River Murray’, Geotechnical Investigation Report. SKM 2010b, ‘Study into river bank collapsing for Lower River Murray: Appendix A Riverfront Road – Murray Bridge in-situ and laboratory test results’, Geotechnical Investigation Report. SKM 2010c, ‘Study into river bank collapsing for Lower River Murray: Appendix B Caloote in-situ and laboratory data’, Geotechnical Investigation Report. SKM 2010d, ‘Study into river bank collapsing for Lower River Murray: Appendix C Woodlane Reserve in-situ and laboratory data’, Geotechnical Investigation Report. SKM 2010e, ‘Study into river bank collapsing for Lower River Murray: Appendix E East Front Road in- situ and laboratory data’, Geotechnical Investigation Report. SKM, 2011, ‘Research into processes, triggers and dynamics of riverbank collapse in the Lower River Murray’. SoilVision, 2009a, SVFlux, ver. 4.23, . SoilVision, 2009b, SVSlope, ver. 4.23, . Springer, FM, 1981, ‘Influence of the rapid drawdown events on river bank stability’. Masters research thesis, University of Louisville. Springer, FM Jr, Ullrich, CR and Hagerty, DJ, 1985, ‘Streambank stability’. Journal of the Geotechnical Engineering Division, ASCE, 111(5), 624–640. Springer, FM, 1981, ‘Influence of rapid drawdown events on river bank stability’, University of Louisville. Stark, TD and Eid, H, 1998, ‘Performance of three-dimensional slope stability methods in practice’. Journal of Geotechnical and Geoenvironmental Engineering, 124(11), 1049–1060.

Page 74 of 77 Riverbank Collapse Task 5 Report

Stark, TD, 2003, ‘Three-dimensional slope stability methods in geotechnical practice.’ Proceedings of 51st Annual Geotechnical Engineering Conf., University of Minnesota, Minnesota Geotechnical Engineering Group, Minneapolis, February, 41–74. Sterrett, R and Edil, T, 1982, ‘Ground-water flow system and stability of a slope’. Ground Water, 20(1), 5–11. Tajeddin, R, Ramsey, N, Pain, D, Mollison, D and Sandercock, P, 2010, ‘Influence of water level risk on river bank collapse, Lower River Murray’. Sinclair Knight Merz, Australia. Tang, WH, Stark, TD & Angulo, M, 1999, ‘Reliability in back analysis of slope failures’. Soils and Foundations, 39(5), 73–80. Thorne, CR and Osman, AM, 1988, ‘Riverbank stability analysis 2: Applications’. Journal of Hydraulic Engineering, ASCE, 114(2), 151–172. Thorne, CR, 1982, ‘Processes and mechanisms of river bank erosion.’ In Gravel-bed rivers, Eds. RD Hey, JC Bathurst and CR Thorne, John Wiley & Sons, Chichester, 227–271. Thorne, CR, Hey, RD, Newson, MD, 1997, ‘Bank erosion and instability’. In Applied fluvial geomorphology for river engineering and management, Eds. CR Thorne, RD Hey and MD Newson, John Wiley, New York, 137–172. Torres Hernandez, G, 2011, 'Estimating the soil-water characteristic curve using grain size analysis and plasticity index'. Masters thesis, Arizona State University. Twidale, CR, 1964, ‘Erosion of an alluvial bank at Birdwood, South Australia’. Zeitschrift für Geomorphologie NF, 8(2), 189–211. Urgeles, R, Leynaud, D, Lastras, G, Canals, M and Mienert, J, 2006, ‘Back-analysis and failure mechanisms of a large submarine slide on the ebro slope, NW Mediterranean’. Marine Geology, 226(3-4), 185–206. Wang, H, Liu, G, Xu, W and Wang, G, 2005, ‘GIS-based landslide hazard assessment: An overview’. Progress in Physical Geography, 29(4), 548–567. Wang, L, Hwang, JH, Luo, Z, Juang, CH and Xiao, J, 2013, ‘Probabilistic back analysis of slope failure – a case study in Taiwan’. Computers and Geotechnics, 51, 12–23. WaterConnect, 2014, ‘Real time water data – Latest hourly observation', Government of South Australia, . WaterConnect, 2015, ‘Riverbank collapse historical incident website', Government of South Australia, . Wen, HJ, Li, X and Zhang, Jl, 2009, ‘An evaluation-management information system of high slope geo- risk for mountainous city based on GIS’. Proceedings of 1st Int. Conf. on Information Science and Engineering (ICISE 2009), Nanjing, China, December 26-28, 1975–1978. Xie, M, Esaki, T and Zhou, G, 2004, ‘GIS-based probabilistic mapping of landslide hazard using a three- dimensional deterministic model’. Natural Hazards, 33(2), 265–282.

Page 75 of 77 Riverbank Collapse Task 5 Report

Xie, MW, Esaki, T and Cai, MF, 2006, ‘GIS-based implementation of three-dimensional limit equilibrium approach of slope stability’. Journal of Geotechnical and Geoenvironmental Engineering, 132(5), 656–660. Xie, MW, Zhou, GY and Tetsuro, E, 2003, ‘GIS component based 3D landslide hazard assessment system: 3DSLOPEGIS’. Chinese Geographical Science, 13(1), 66–72. Yan, ZL, Wang, JJ and Chai, HJ, 2010, ‘Influence of water level fluctuation on phreatic line in silty soil model slope’. Engineering Geology, 113(1-4), 90–98. Zaitchik, BF and Van Es, HM, 2003, ‘Applying a GIS slope-stability model to site-specific landslide prevention in Honduras’. Journal of Soil and Water Conservation, 58(1), 45–53. Zapata, CE, Houston, WN, Walsh, KD and Houston, SL, 2000, ‘Soil–water characteristic curve variability’. Advances in Unsaturated Soil Mechanics, Geotechnical Special Publication No. 99, eds CD Shackleford, SL Houston and N-Y Chang, Geo-Institute of ASCE, 84-124. Zhang, JF, Li, ZG and Qi, T, 2005, ‘Mechanism analysis of landslide of a layered slope induced by drawdown of water level’. Science in China Series E Engineering & Materials Science, 48(1), 136– 145. Zhang, LL, Zhang, J, Zhang, LM and Tang, WH, 2011, ‘Stability analysis of rainfall-induced slope failure: A review’. Proceedings of the Institution of Civil Engineers-Geotechnical Engineering, 164(5), 299– 316. Zolfaghari, A and Heath, AC, 2008, ‘A GIS application for assessing landslide hazard over a large area’. Computers and Geotechnics, 35(2), 278–285.

Page 76 of 77 Riverbank Collapse Task 5 Report

The Goyder Institute for Water Research is a partnership between the South Australian Government through the Department of Environment, Water and Natural Resources, CSIRO, Flinders University, the University of Adelaide, the University of South Australia and ICE WaRM (the International Centre of Excellence in Water Resources Management).

Page 77 of 77