Research Topics

GIS-based Landslide Inventory Sydney Basin

The University of Wollongong GIS-based Landslide Inventory is currently being expanded from its Illawarra centric coverage (700 landslides) to include the 31,000km2 geological extent of the Sydney Basin and ultimately all of New South Wales (hereafter this inventory will be referred to as the NSW LI). In 1998, this inventory stood at 319 sites of instability and in 2010 it had grown to 616 landslides. When this current phase of expansion is completed, the NSW LI will include up to 2000 landslide sites, and it will continue to expand over the next few years. The NSW LI is being redesigned following an international literature review and is being re-compiled into an ESRI ArcGIS v10 Geodatabase (Flentje et. al., 2012).

This project will develop Landslide Susceptibility zoning maps of a scale and resolution to potentially form the basis for land use planning and the generation of development control areas for Local Government use. Furthermore, these maps will be produced in a consistent and transparent manner across the wider Sydney Basin region. This proposal will complete this work within a Geographic Information System (GIS) data management environment using data mining techniques.

Research aim and objectives

Develop Landslide Susceptibility zoning maps that will potentially form the basis for land use planning and the generation of development control areas for Local Government use.

1. To compile a GIS – based landslide inventory for wider Sydney basin area

2. To build a python script/VB.NET add-in to integrate the GIS and data mining techniques and automate the process of data preparation, data mining and converting the data mining outcome in to a raster map.

3. To develop a landslide susceptibility model

Developments

Arc Map Tool Bar Add-In has been developed to integrate the GIS and data mining techniques and automate the process of data preparation, data mining and converting the data mining outcome in to a raster map.

Refined landslide inventory structure to facilitate the better storing and querying of the landslide data.

A brief overview of landslide monitoring

Landslide monitoring is required for a wide variety of reasons. These may include; to determine the extent, magnitude and style of landslide movement, for risk and even emergency risk management assessments and/or to assist with the design and implementation of site remedial and/or mitigation works.

Landslide monitoring can be undertaken in many forms. The monitoring may include traditional ground survey techniques repeated over time to determine movement at specific locations. It may include subsurface investigations with the aid of borehole inclinometers and their periodic manual monitoring and even continuous automated monitoring (as discussed below). Over recent years, automated robotic ground survey using various laser scanning techniques have been shown to be cost effective in certain applications. Even satellite based radar interferometry techniques over large areas at moderately high grid resolutions, with movements being detected down to cm level resolutions are being applied.

Inclinometer Monitoring

Where subsurface investigations are being carried out a widely used technique internationally involves the installation of inclinometer casings in boreholes and subsequent monitoring of the casings (typically installed vertically) using manual inclinometer probes. This technique is summarised in Figure 1.

Figure 1. Principle of Inclinometer operations, after Green and Mikkelsen, 1988.

With time, repeated visits to the site whereby full length casing profiles can be recorded, as shown in Figure 2, the displacement history of the full subsurface profile can be determined. In Figure 2, the ground surface is shown as the top most horizontal line and depth below the ground surface is shown on the left hand y axis in meters. Horizontal displacement of the ABS plastic inclinometer casing is shown in the bottom horizontal x axis in millimetres. The data shown in Figure 2 has been recorded at a landslide in the northern suburbs of Wollongong and the borehole drilling at this location shows bedrock occurs at a depth of approximately 8.2m. The inclinometer profiles clearly show landslide subsurface shear displacement occurring at a depth of between 7m and 8m below ground level with approximately 150mm of shear occurring at a depth of 7m below the ground surface. The first profile shown in this graph (the vertical y axis) was recorded on the 9th March 1989, and the last profile shown in Figure 2 was recorded on the 15th May 1996.

Figure 2. Inclinometer profiles recorded at a landslide in Scarborough, in the northern suburbs of Wollongong.

Such manual inclinometer profiling techniques have been employed in the Wollongong area for up to three decades. Since the mid 1990's, the UoW LRT has employed Coffey Geotechnics to conduct inclinometer monitoring at a number of landslide locations around the Illawarra area.

Monitoring of pore water pressure

Ground water and pore water pressure plays an important role in landsliding. Rising pore water pressures, often (but not always) the result of rainfall, are typically the triggering mechanism for landsliding in the Illawarra region. Man made effects are another common triggering mechanism. Therefore, monitoring ground water pressure is an important component of any landslide monitoring strategy.

There are many ways to measure ground water and specifically pore water pressures. Currently the UoW LRT uses vibrating wire piezometers installed in sealed and isolated narrow sand filled intake zones at or near the landslide surface to monitor pore water pressures.

Rainfall Monitoring

Due to its influence of ground water and pore water pressure, rainfall is an important component of any landslide monitoring strategy. The UoW LRT currently employs tipping bucket style Pluviometers to monitor rainfall.

Monitoring History

The manually recorded monitoring history from a Scarborough landslide is shown in Figure 3. Daily rainfall totals from of Bureau of Meteorology daily rain gauge is shown along the base plot as a histogram. The cyclic periods of wet and dry are shown with the cumulative rainfall curves which total daily rainfall for the preceding 7, 30, 60, 90 and 120 day periods. Overlying this are two curves of total cumulative shear displacement as measured in two different borehole inclinometers. Above that is the really interesting section of this graph which shows the rate of landslide shear, the actual manually monitored rate of landslide movement.

Figure 3. Site 64 Monitoring Jan 89 to end 99

This style of monitoring was discussed in detail by Flentje during his PhD research. Flentje's PhD thesis (particularly in Chapters 8 and 9) can be downloaded .

Whilst such monitoring is very useful in geotechnical engineering and engineering geological assessment of landslides, several issues are very clear. Manual monitoring is expensive, and is only undertaken on an 'as needed' basis. The stepped nature of the cumulative displacement curve in Figure 3 reflects not only the timing of landslide movement, but also the timing of landslide monitoring. Provided the monitoring occurs frequently, periods of movement are detected. If the monitoring is undertaken at low frequencies of say, annually, or even less frequently, episodes of movement may be lost in the 'averaging' effect of the long return period of the monitoring. Furthermore, such manual monitoring is not 'pro-active', it only tells you what has happened and requires some considerable reporting effort to get the information to management.

Continuous Landslide Monitoring

Following a long period of manual monitoring of landslides in the Wollongong region, the UoW LRT has undertaken the development of a network of Continuous Real-Time Monitoring (CRTM) landslide field stations.

Using all of the afore mentioned techniques, we have simply replaced manual inclinometers, with In Place Inclinometers (IPI's) which actually hang in borehole casings, at designated depths where manual inclinometer profiling has confirmed shear planes, or in the absence of confirmed shear planes, where it is assumed the shear planes have developed. For example, in Figure 2 above, we would position an IPI to monitor shear displacement over the interval 7 to 8m as we know the style of landslide displacement in this borehole and we would expect this style of displacement to continue.

Using solar power and batteries, we have installed data loggers out on various important landslide sites and connected to the loggers a series of IPI's, vibrating wire piezometers, extensometers, tilt meters and rain gauges to monitor various aspects of landslide and other structural movement.

The continuously monitored history of a landslide in the northern Wollongong suburbs is shown in Figure 4. The period covered by this graph is late 2003 to early 2005 and the data shown is recorded hourly. Figure 4 includes a daily rainfall histogram along the bottom which shows two events with daily totals exceeding 80mm per day in April and October 2004. Over this period the In Place Inclinometers have provided excellent data. The IPI rate of shear displacement curves clearly displays two prominent spikes of accelerated displacement commencing on the 4th April and the 21st October 2004. The movement event, which commenced on The 4th April, continued for 5 days and peaked at 2.4mm per day. This was triggered by rainfall of 110mm and 106mm on consecutive days. The movement event, which commenced on the 21st October 2004, lasted for 7 days and peaked at 2.7mm per day on the second day. A maximum daily rainfall of 81.5mm and several other days of 15mm to 30mm triggered this short duration of movement.

The vibrating wire piezometers (vwp) have both shown quite strong responses to the rainfall.

Figure 4. Continuously logged monitoring history of a northern suburbs landslide for the period late 2003 to early 2005.

This style of monitoring data is available in near real-time (subject only to the on site logging interval) via a secure password protected web-interface developed by the UoW LRT and the UoW CEDIR group. The site is password protected as it is a research application at this stage and it is constantly evolving.

The information being collected is enhancing our understanding of landslide and triggering mechanisms. The information is greatly facilitating the quantitative assessment of landslide frequency.

Landslide triggering rainfall thresholds

Rainfall is recognised nationally and internationally as a major triggering factor for the initiation of slope instability and the initiation of landslide movement. Such movement may range from subtle, minor displacements to catastrophic scale movements in terms of velocity and travel distance. Extensive global research efforts have focused on the development of rainfall thresholds for the initiation of slope movements. However, an important source of uncertainty in hazard and risk assessment comes from both the spatial and temporal variation of rainfall (Chowdhury and Flentje, 2002).

In so called rainfall triggered landslides, pore water pressure is the actual trigger of the onset of the landsliding. However, measuring pore water pressure is complex and difficult and, generally speaking, rarely undertaken, at least historically. In comparison, however, the monitoring of daily rainfall is widespread internationally, and the recorded rainfall data is, in general, widely accessible. Furthermore, there is a clear conceptual relationship between rainfall and pore water pressure, albeit considerably more complex than the simple concept implies. Therefore, the relationship between rainfall and the onset of landsliding has been widely studied.

At the University of Wollongong significant research has been directed to the analyses of rainfall, pore water pressure and slope movements enabling estimation of such thresholds and the associated uncertainties. Our research relates to the periodic and continuous monitoring of landslides, the development of a GIS-based inventory of landslides, GIS-based analyses of selected rainfall events and it is also supported by others internationally (Caine 1980, Wieczorek 1987). The LRT confirms that the shorter duration thresholds (6 hours to 3 days for this study area) are most relevant for shallow debris flows and shallow slides whilst the longer duration thresholds (up to 90 days for this study area) are most relevant for deeper seated slide and slide-flow category landslides. It is important to note that there is little research concerning triggering rainfall for the initiation of rockfalls within the study area.

Overview of Wollongong Regional Rainfall

Average Annual Wollongong Rainfall

Wollongong has a cool temperate climate with an annual average rainfall of approximately 1200mm across the coastal plain. The orographic effect of the escarpment on rainfall is quite pronounced, as shown in Figure 1. The annual rainfall is closer to 1600mm on the higher ground immediately to the west of the escarpment south of Bulli, and approximately 1500mm on the intermediate to upper escarpment slopes (Young 1976, Mills and Jakeman 1995).

Figure 1. Average annual rainfall based on Bureau of Meteorology records.

Intensity Frequency Duration of rainfall in the Wollongong area

The historical frequency and duration of rainfall intensity determined from the rainfall record at any rainfall station can be graphically presented as Intensity Frequency Duration or IFD plot. The average, or expected, value of the periods between exceedances of a given rainfall total accumulated over a given duration, or the Average Recurrence Interval (ARI), for 14 durations ranging from 3 hours to up to 30 days has been determined for one location in Wollongong as shown in Figure 2. ARI curves have been prepared for 1, 2, 5, 10, 20, 50 and 100 years. This rainfall analysis was completed for the University of Wollongong by the Bureau of Meteorology (BOM) Hydrometeorological Advisory Service (HAS) one year after the August 1998 rainfall event.

Whilst this rainfall analysis is specific to the Woonona Rainfall Station 068108 (Latitude 34.350 S Longitude 150.900 E) it serves as a useful guide across the study area, particularly at similar lower escarpment elevations.

Figure 2. BOM HAS Daily Rainfall Intensity Frequency Duration Analysis for Woonona Rainfall Station 068108.

Background

For rainfall-triggered landslides, researchers often present the triggering rainfall as a relationship between rainfall intensity and duration. For example, Caine (1980) reported on the relationship between threshold rainfall intensity and duration on the one hand and the occurrence of shallow landslides and debris flows on the other. Caine summarised the data concerning a selection of 73 international reports of shallow landsliding and proposed a landslide triggering threshold rainfall curve in the form:

I = 14.82 x D^-0.39

where I is the rainfall Intensity in mm per hour and D is the duration of rainfall in hours. This curve is shown in Figure 3. Any combination of I and D that plots below this curve will not trigger landsliding. A comprehensive relationship of this type had not been developed in Australia until recent UOW research and the analysis of August 1998 rainfall that is discussed later in this report.

For landsliding within the Wollongong region, several authors have reported on landslide triggering rainfall magnitudes. Bowman (1972) concluded that catastrophic landslides occur after rainfall of over 430mm in 1 month, and slides often occur after falls over 350mm per month. Young (1976) proposed that a lower critical value of 250mm rainfall per month was likely to initiate landslip. Longmac Associates (1991, reported in Pitsis 1992) found a poor correlation between landsliding and one month antecedent rainfall totals. They concluded that a three monthly period of antecedent rainfall correlated better with the occurrence of large-scale landslides within the area between Stanwell Park and Coledale. In a geotechnical report for a large landslide affecting Morrison Avenue in Coledale (site 77), Longmac Associates suggested a three month threshold between 550mm and 650mm to trigger landsliding. In another report for a landslide in Scarborough, site 64, Longmac Associates suggested that a one month rainfall total exceeding 350mm was required to initiate landsliding, and a one month total of 250mm was required for the continuation of the movement after first initiation.

Clearly, this is a complex issue and there is unlikely to be one simple answer. This is even more complicated given the variety of landslide types, and landslide volumes that are known to occur within the Wollongong region.

Landslide Inventory

Regional versus site specific thresholds

Two types of landslide triggering rainfall thresholds may be considered, namely regional thresholds and site-specific thresholds. Regional thresholds covering a wide variety of landslide scenarios would, in general, be lower than site-specific thresholds. Continuous real-time monitoring (CRTM) has been proceeding at some landslide sites in Wollongong from early 2003.

 

Such monitoring provides reliable data to establish site-specific landslide triggering rainfall thresholds. A Natural Disaster Mitigation Project supported through the Federal Department of Transport and Regional Services (DOTARS) and the State Office for Emergency Services (SOES) with RailCorp and the University of Wollongong is enabling this developing technology to be applied at more sites. Such detailed applied research will enable the assessment of accurate site-specific landslide triggering rainfall thresholds. The UOW has been researching both site specific and regional thresholds using two general methods.

One method considers cumulative periods and respective magnitudes of rainfall for these periods in relation to the timing of the onset of slope movements as determined by reports or periodic observation, including inclinometer monitoring. The second method considers triggering rainfall intensity and duration in relation to the locations of landsliding triggered by the given rainfall event within the region. The August 1998 rainfall event which affected the Wollongong region was an extreme and widespread rainfall event and it triggered a large number of landslides. Post 5pm on the 17th August and in the weeks following the event, 142 landslide locations were recorded and each site was inspected and mapped (GTR, 1998)This event is discussed in some detail in the following sections.

Overview of the 5 day event 15th - 19th August 1998

The rainfall event extended from 6 p.m. Eastern Standard Time (EST) on Saturday 15 August to 6 p.m. EST on Wednesday 19 August 1998. It is important to recognise and assess both the spatial and temporal variation of rainfall. To facilitate this analysis of this rainfall event, the recorded rainfall data has been analysed extensively with the aid of GIS whereby a series of interpolated 10m x 10m grids have been developed based on the recorded rainfall at each rainfall station.

The highest rainfall total over this period of 745mm was recorded at the RTA pluviometer rain gauge on the Mt Ousley road section of the F6 freeway. Extremely heavy rainfall occurred on 17 August 1998 between 5 p.m. and 8 p.m. causing flash flooding with extensive damage to property and the loss of one life. Up to 5 p.m. on 17 August the rainfall total was 375mm over 47 hours. From 9 am Monday 17 August to 9 am Tuesday, 18 August the rainfall total was 445mm. The temporal pattern of the rainfall event and the accumulated rainfall from Mount Ousley rainfall station is shown in Figure 3.

Figure 3. Temporal pattern of 15 minute and cumulative rainfall at Mount Ousley

The intensity-frequency-duration curves derived from pluviometer rainfall monitoring stations around the city show that for intervals between 3 hours and 12 hours, some stations exceeded a 1 in 100 year event, and rainfall totals at one station for an 8 hour interval exceeded a 1 in 200 year event. Average recurrence intervals exceeding 1 in 100 years were determined for durations between 30 minutes and 24 hours along the top of the escarpment near Mount Ousley, Rixons Pass and Bulli Pass (Evans and Bewick, 1999) as shown in Figure 4. However, many parts of the city experienced significantly less falls over a wide range of durations equating to say a 1 in 20 year event for a 24 hour period.

Figure 4. Wollongong City Intensity Frequency Duration curves (from 1 to 500 years) with selected August 1998 storm event maximum intensity frequency duration plots. Pattern of 15 minute and cumulative rainfall at Mount Ousley

The spatial distributions of cumulative rainfall over different antecedent have been analysed with the aid of the GIS. The antecedent time periods of 6 and 12 hours prior to 7pm on the 17th August and 1, 3, 5, 7, 30, 60, 90 and 120 days prior to 9.00am on the 17th, 18th and 19th August have been considered in the various analyses (Grootemaat 2000, Murray 2001). Two examples of the spatial distribution of rainfall the August 1998 rauinfall event are shown in Figure 5 and 6. Figure 5 shows mapped August 1998 landslides and 12hr rainfall (mm) to 9.00am 18th August 1998 while Figure 6 shows mapped August 1998 landslides and 10 day rainfall (mm) to 9.00am 19th August 1998.

Figure 5. Mapped August 1998 landslides and 12hr rainfall (mm) to 9.00am 18th August 1998

Figure 6. Mapped August 1998 landslides and 10 day rainfall (mm) to 9.00am 19th August 1998

A regional intensity frequency landslide triggering rainfall threshold for Wollongong based on the August 1998 event. Using the GIS-based raster rainfall distributions, the rainfall magnitude at the centre of each landslide for each cumulative period, including those shown in Figure 5 and 6, has been recorded for all 142 sites activated during the August 1998 event. This data has been compiled into an excel spreadsheet and plotted as shown in Figure 7. Figure 7 shows the rainfall magnitudes for each antecedent rainfall period (one bar for each period) as a series of 142 crosses (one cross for each landslide site per vertical bar) making up each vertical bar. The curve extending across the graph near the base of each vertical bar is, therefore the regional threshold for the city of Wollongong for the August 1998 event.

Figure 7. The lower bound regional landslide triggering rainfall threshold for the city of Wollongong during the August 1998 event (red curve).

Examples of Site Specific Thresholds

The sites discussed in the following section are amongst 142 known locations where landslide movement was triggered during the 15-19th August 1998 rainfall event.

Site 340, the Mt Kembla Spur landslide, early morning 20th August 1998

Landsliding at site 340 was first reported in the early hours of 20th August 1998 when local residents of Mt Kembla village heard loud noises emanating from the dark escarpment above the village. An inspection by helicopter later that morning confirmed the landslide location and extent. The noise emanating from the landslide site was the sound of flowing water and of large 30m high Eucalyptus trees and debris including boulders falling off a 10-15m high Sandstone cliff face and landing on a near flat terrace adjacent to the fire trail. At the time of the walkover inspection, late on the morning of the 20th August, the landslide was still active and debris was still moving over the cliff face. The landslide includes a large deep-seated translational slide type of failure on the Stanwell Park Claystone terrace with a volume of approximately 50,000m³ and the debris flow material of approximately 2000m³. The rainfall interpolated by the GIS based analysis for this site is summarised in Table 1 and displayed graphically in Figure 8.

Table 1. Cumulative rainfall to 9.00am 20th August 1998 interpolated by GIS-based analysis to the centre of site 340.

Site 343, the Mt Keira Ring Track landslide, evening of the 17th August 1998

Landslide movement at site 343, occurred at about 7pm on the 17th August 1998. The landslide is a slide-flow category landslide having a volume of approximately 3500m³. Sliding occurred and soon developed into a flow on the steep lower slopes of the site. The flow component of this landslide inundated a culvert on the Kemira Colliery platform.

The rainfall interpolated by the GIS based analysis for this site is summarised in Table 2 and displayed graphically in Figure 9.

Figure 9. Cumulative rainfall IFD curve to 7.00pm on 17th August 1998 indicative of movement at Site 343 at this time.

Site 393, the Bulli Pass debris flows, 5pm to 7pm on the 17th August 1998

By 7pm on 17th August 1998, Bulli Pass was closed to traffic by the State Emergency Services and a number of cars inundated with debris were trapped in the upper section of the road as shown in Figure 10. An aerial view of the upper Bulli Pass slopes is shown in Figure 11. While the slopes were too inaccessible and steep to map in detail at close range, it has been estimated that approximately 30 debris flows occurred, each with an average volume of approximately 200m³. This area has been mapped as one debris flow site, Site 393 in the UoW Landslide Inventory.

Figure 10. The upper section of Bulli Pass, photo taken 7.30am 18th August 1998 by SES.

Figure 11. Bulli Pass from the Polair 2 during the afternoon of the 20th August 1998.

The GIS-based interpolated rainfall for this event are shown in Table 3 and plotted graphically in Figure 12.

Table 3. Cumulative rainfall to 7.00pm 17th August 1998 interpolated by GIS-based analysis to the centre of site 393.

Figure 12. Cumulative rainfall IFD curve to 7.00pm on 17th August 1998 indicative of movement at Site 393 at this time.

Based on these types of investigations, the UoW LRT has proposed the landslide triggering rainfall threshold as shown as the solid dotted line in Figure 13. This is a regional type threshold, above which disruptive type landslide movement can be expected to occur within the Wollongong region. Again, this is superimposed on the IFD curves for Woonona rainfall station 68108 so that the reader can gain an appreciation of the frequency of occurrence of these levels of rainfall. The red and blue curves on Figure 13 have been derived from continuous monitoring of a landslide in the northern suburbs during the 3 year period 2004 to 2006. The blue curve shows the rainfall intensity duration relationship where the monitored landslide moved at up to 5mm during a rainfall event. The red curve shows the rainfall intensity duration relationship where the monitored landslide moved at up to 10.5mm during a rainfall event. Landslide movement of greater magnitudes has not been recorded during the period of continuous monitoring.

Figure 13. Station 68108 IFD curves superimposed upon with UoW LRT landslide triggering threshold.

The two lower curves are based on the continuous monitoring record and show rainfall thresholds for two different magnitudes of movement. These curves are preliminary for one site only and are based on the work of Bellew (2006).

References

Bowman, H. N., 1972a. Natural Slope Stability in the City of Greater Wollongong in the Records of the Geological Survey of New South Wales Volume 14, Part 2, 29/9.

Chowdhury, R. and Flentje, P., 2002. Uncertainties in Rainfall-Induced Landslide Hazard. Quarterly Journal of Engineering Geology and Hydrogeology. Symposium in Print on Landslides. London (UK), Volume 35 Part 1, February 2002, pp 61-70.

Caine, N., 1980. The Rainfall Intensity Duration Control of Shallow landslides and Debris Flows, Geografiska Annaler, Vol 62, A 1-2, pp 23 Ð 27.

Geotechnical Team Report (GTR), 1998. Geotechnical Services Emergency Storm Event August 1998, for the Wollongong City Council (Unpublished).

Grootemaat, G, 2000. Spatial Characteristics of High Magnitude Rainfall Events and Channel Response in the Illawarra Region. B.Sc. Honours thesis, School of Geosciences, University of Wollongong, unpublished. 154 pages.

Mills, K and Jakeman, J., 1995. Rainforests of the Illawarra District. Coachwood Publishing, Jamberoo, 143 p.

Murray, E, 2001. Rainfall Thresholds for Landslide Initiation in the Wollongong Region, Internal report to Australian Geological Survey Organisation and SPIRT Project Team at the University of Wollongong.

Wieczorek, G.F., 1987. Effect of rainfall intensity and duration on debris flows in central Santa Cruz Mountains, California. In: (Costa and Wieczorek editors) Debris flows/avalanches: processes, recognition and mitigation. Reviews in Engineering Geology, Geological Society of America, 7:23-104.

Young, A. R. M., 1976. The distribution, characteristics and stability of debris mantled slopes in northern Wollongong. Master of Science thesis, University of Wollongong (unpublished).

Landslide Hazard And Risk Management

Important state of the art documents pertaining to Landslide Risk Management in Australia have been published by the Australian Geomechanics Society in both 2000 and 2007. A link to these documents is provided elsewhere on this web page under the Menu item .

Both site specific and regional assessments are important for landslide hazard and risk. These are discussed separately after considering the basic terminology.

The basic terminology

The terms hazard and risk are used in many areas of life and human endeavour such as the natural environment, trade and investment, economic development, human health and safety, engineered structures, transport and other infrastructure, damaging events (rainstorms, floods, earthquakes, cyclones etc.). A characteristic feature of hazard and risk in any particular application is, of course, uncertainty with regard to the contributing factors or causes. The definition and usage of the terms may differ from one application to another. Therefore, it is appropriate to limit attention here to slopes, earth structures and associated phenomena such as landslides. For slopes and landslides, it is important to distinguish between first occurrence and reactivation or renewal of movement. The hazard and risk may be quite different for the two different failure stages or phenomena.

Hazard may be regarded as a threat of economic loss, and/or a threat to human safety (death, injury) and/or a threat to the quality and safety of the environment. The most important aspects of hazard are the probability of its occurrence (also called likelihood) and its magnitude. As stated earlier, the dominant feature is uncertainty. In relation to slopes, there may be uncertainty relating to location of failure occurrence (where?), the timing of occurrence (when?), the magnitude (how large, how much deformation or movement or travel distance?), the velocity of the event (how fast?) and the frequency (how often?). The terms "failure susceptibility" or "landslide susceptibility" are often used to denote the probability of occurrence of a slope failure or landslide. In connection with landslides, the term "landslide susceptibility" is often used.

It may be important to distinguish between primary hazard event and secondary hazards, follow-on hazards and post-event hazards. For instance an earthquake (primary hazard) may trigger one or more landslides (secondary hazards). The debris from a significant landslide (primary hazard) may block a river or stream forming a landslide dam (secondary hazard). The basic questions concerning a potential hazard of slope failure or landslide are:-

• What is the probability?
• How large is the potential failure?
• What kind of mechanism of failure?
• Will the deformation be limited and slow or will it occur catastrophically?
• What is the potential travel distance?

Risk may be regarded as the potential for adverse consequences related to particular hazard or hazards. It is therefore necessary to consider what assets and elements are at risk from a slope or landslide hazard and to assess the value and vulnerability of each of those assets. The first basic question with regard to risk concerns the expected economic loss. In this regard, quantification of risk involves the product of probability of event and the expected value of loss and damage which is the product of value and vulnerability of the asset.

RISK	= Probability x Adverse Consequences
	= Probability x Expected Loss and Damage

The second basic question with regard to risk is:

• "What is the probability of loss of human life?"

This probability is a product of the hazard probability, the impact probability and the vulnerability of individuals.

Vulnerability of a particular asset or element is the degree to which harm or damage will occur to that asset or element. This may be expressed as a factor varying from 0 to 1, the same limits which apply to probability. Assessment of vulnerability requires consideration of a number of factors and the choice of values is likely to be dominated by subjective judgment.

Types of Risk

As stated earlier, one must distinguish between risk to life and safety, economic risk and environmental risk. Moreover, it is also important to distinguish between individual risk and societal risk and between voluntary and involuntary risk. In connection with slopes and landslides, it is important to distinguish between site-specific and regional risk. It is also essential to describe fully how the risk has been determined and whether the outcomes are qualitative, semi-quantitative or quantitative assessments.

Definition of categories for site-specific assessments

Definitions of likelihood (hazard), consequence and risk were proposed, among others by AGS 2000 and 2007. According to Ko Ko et al. (2001), hazard level of sites may be characterized in the following 5 hazard (or likelihood ) categories with the annual probability of failure shown in parenthesis: (a)Very High (>0.2), (b) High (0.2-0.02), Medium (0.02-0.002), Low (0.002-0.0002) and (e) Very Low (0.0002). Degrees of consequence may also be defined in specific categories. Risk is defined in terms of matrix of likelihood and consequence.

Qualitative assessment based on site inspections

During a site inspection, different attributes of a slope may be allocated relative scores or weights based on experience and judgment. This exercise may be done separately for likelihood and consequence (separately for economic and human safety). Based on total scores for each of likelihood and consequence, risk level may be determined. The procedure is different in some respect for each type of slope (e.g., natural slope, embankment, excavation etc.). In this way, a consistent guide to a qualitative, site specific, assessment of hazard and risk for different types of slopes has been outlined by Ko Ko (2001) and published in summary by Ko Ko et al (2004). An effort has been made to validate the results of such studies for several sites within a study area. There are other qualitative approaches developed by relevant organizations for their internal use and some of these may have been published in the literature.

Quantitative approach - acceptable or tolerable levels of risk

In cases where comprehensive engineering geological site investigations and or detailed geotechnical analysis has been carried out, and where likelihood, consequence and element vulnerability can be quantified, the risk levels may be estimated quantitatively.

The perception of risk amongst individuals and groups varies widely depending on the type or source of the risk as well as on the level of awareness and exposure. Most people have little or no personal exposure to risks associated with the failures of engineered structures such as major buildings, bridges and dams. The situation is similar with regard to natural slopes, embankments and excavated slopes. Even in urban areas affected by periodic landsliding, the understanding and perception of landslide risk to individuals may vary widely.

One may start with a general principle concerning acceptable or tolerable risk. The level of risk from a project or facility to an individual or society which can be accepted or tolerated should be inversely proportional to the adverse consequences of failure. However, the level of acceptability or tolerance of risk also depends on risk awareness or perception and also on whether the risk is voluntary or involuntary. Guidelines can be found in the literature. The individual risk criteria proposed by ANCOLD (1994) concerning dams include the following:

• As an average over the population, an objective risk level of 1 in a million (1 in 1,000,000) per exposed person per year
• For the person most at risk, an objective risk level of 1 in hundred thousand (1 in100,000) per exposed person per year
• For existing dams, a limit of 10 times the objective values subject to the principle of ALARP (as low as reasonably practicable). Risks should be regarded as tolerable only if further risk reduction is impracticable or if the cost is grossly disproportionate to the benefit.

Regional Assessment of Hazard and Risk

Introduction

Regional studies for landslide hazard and risk assessment are extremely useful for the development and implementation of risk management strategies. Such studies also facilitate the planning of the extent and intensity of development in sloping areas. The results of such studies are often consolidated and presented as maps of landslide susceptibility and hazard. In some cases risk maps may also be developed.

The processes involved include:

• The development of a Landslide Inventory
• Assessment of landslide frequency
• Assessment of Landslide Hazard
• Assessment of the consequences of landsliding
• Assessment of element vulnerability
• Risk Assessment whereby risks are determined and assessed with respect to assessment criteria (tolerable/acceptable levels, the ALARP principle, economic implications and budgets etc.
• Risk Management.

The development of GIS (Geographical Information Systems) applications has revolutionized the processes and methods of analysis by which the development of such zoning are developed. GIS is a versatile tool which facilitates the, collection, organization, verification, quality control, analysis and synthesis and display of data or information concerning a particular region or phenomenon. Often the final maps are obtained from a synthesis of disparate input data, representing the spatial distribution of each contributing or influencing factor such as geology, existing landslides, slope inclination, significant geotechnical properties, vegetation, moisture, extent of development, elements at risk from landsliding, etc. GIS-based methods can handle considerable volume and variety of spatial and temporal data at quite high resolutions. More importantly, GIS based analyses and maps can be updated frequently and conveniently in the light of additional data and new information.

Qualitative and quantitative approaches for regional analysis

Regional susceptibility and hazard assessment methods may be qualitative or quantitative. However, the way in which methods are classified is somewhat subjective and some methods may be best described as semi-quantitative. Of course, a quantitative approach requires availability of good quality information and data and minimal reliance on subjective judgment. For every method, qualitative or quantitative, one must ask the question as to whether relative or absolute susceptibility and hazard are being assessed. The final outcomes are presented invariably as maps of hazard and susceptibility showing zones in a number of categories depending on the needs of the project and the project goals. For example, the degree of hazard and susceptibility may be classified in five categories as "very high", "high moderate", "low" and "very low". If a quantitative method has been used, it may be useful to specify the magnitude of likelihood associated with each of these descriptions.

Qualitative Methods

Fully qualitative methods rely mostly on subjective judgment. The input data may be based on assessments during field visits and, in some cases, supported by aerial photo interpretation. Field geomorphological analyses fall into this category. In such methods, the geoscientist carries out an assessment and/or zonation during site inspection based on expertise and previous experience preferably that obtained in regions similar to the study area in question. The final maps for landslide susceptibility and hazard would then be obtained directly from such field assessments. Of course, the susceptibility and hazard assessed are relative rather than absolute.

Another qualitative approach consists of combining or overlaying of different index maps which are available for the study area or which have been specifically prepared for the qualitative assessment. Such synthesis or combination of information from index maps (for example, slope map, geology map, geomorphic map and land use map) may be carried out with or without weighting of the influencing factors. In any case, if weights are attributed, the basis will be purely subjective. Again the final maps will display relative rather than absolute susceptibility and hazard.

Quantitative methods

Quantitative methods include statistical analyses (bivariate statistical analyses or multivariate statistical analyses), neural networks, knowledge based approaches and, of course, geotechnical engineering approaches (deterministic or probabilistic). A factor of safety approach for a regional study can be easily facilitated within a GIS framework. For example, the "infinite slope" model is often very helpful for the analysis of slopes and especially for assessing the potential for shallow sliding. Just like any other data concerning the study area, the important parameters in an "infinite slope" model (slope inclination, slip surface depth, soil unit weight, pore water pressure, shear strength parameters, etc.) can be attributed to individual pixels within a GIS model for the study area. Thus a factor of safety map can be developed as the susceptibility map or as the basis for such a map.

Similarly, with a GIS framework for a regional study, it is feasible to use a simple probabilistic approach associated with a geotechnical model such as the "infinite slope" model.

However, it is also important to highlight the fact that other approaches such as knowledge-based methods can provide a different kind of insight and may prove to be more advantageous than geotechnical engineering approaches. The latter are, after all, more suited for site-specific studies, having been developed for that purpose. A brief reference is made below to a knowledge-based approach developed for an urban area subject to periodic landsliding.

UOW LRT landslide susceptibility and hazard in the Wollongong region

Research concerning landslide susceptibility and hazard has been in progress since 1993 in the Illawarra region of the state of New South Wales, Australia. The Illawarra region is dominated by an erosional escarpment and it is also a costal region. Initial studies included the development of a geological data base and a comprehensive landslide inventory. In addition to maps of geology and existing landslides, detailed analyses of rainfall have been carried out. Both qualitative and quantitative studies have been carried out for assessing landslide susceptibility and hazard.

Recently, a quantitative, knowledge-based, modelling approach has been used for assessing the hazard and risk. Given the comprehensive landslide inventory which has been developed over more than a decade, the application of such a detailed quantitative approach can certainly be justified. However, considerable effort was required to model the topographic, geological and geomorphological aspects of the study region. Data sets from existing landslide areas as well as from non landslide areas are compared to learn about the similarities and differences and to express these in quantitative terms. The particular knowledge-based method used in this case is generally known by the term data mining. By comparing the detailed characteristic of both landslide areas and non landslide within the study region, rule sets are developed which enable the susceptibility across the whole area to be determined (Landslide Susceptibility) in quantitative terms. The results of the analysis can then be validated by examining their consistency. Moreover, field validation may be carried out in different ways.

Maps of landslide susceptibility and hazard have been developed on the basis of this quantitative approach. The following table (Table 1) shows the categories of susceptibility and hazards which were established (). It is interesting to note that, in the first place the description of hazard is qualitative although the method used to identify the zones is fully quantitative. For regional landslide susceptibility and hazard studies, this is not unusual. In fact, most studies around the world identify and demarcate zones of different degrees of landslide susceptibility and hazard in qualitative terms. However, as is pointed out in the next paragraph, it is possible to quantify these categories relative to each other. It is difficult, if not impossible to express risk quantitatively in absolute values.

Table 1. Hazard zone description, percentage of area affected and percentage of total landslide population in each zone, Illawarra region study, NSW, Australia (Flentje et al, 2007).

Using information from the knowledge-based analysis and from the comprehensive landslide inventory, a quantitative evaluation has been made of the likelihood associated with each zone. Thus we have a probability or likelihood associated with each description. However, it must be recognized that these are relative rather than absolute values. The reason is that much of the information is derived from data including frequencies of landsliding over a limited historical period. Consequently the derived likelihoods can not be regarded as absolute probabilities. Alternative techniques using landslide size frequency models are currently being developed to assess the frequency of landsliding within each of the Susceptibility/Hazard zones.

Preliminary Sydney Basin Wide Susceptibility Zoning

To further investigate and validate the Data Mining methodology that has been completed for a trial area within the Wollongong Local Government area, it is currently being applied to the entire Sydney Basin region and we are currently assembling refined higher resolution data sets to further refine this modelling process. At 25m pixel resolution, this involves a GIS-based raster grid of approximately 220 million pixels not withstanding that this is at a resolution significantly coarser than is preferable. Using similar data sets to those used for the Wollongong study (at 25m pixel resolution), a preliminary landslide Susceptibility model (proof of concept ONLY) has been developed for the entire Sydney Basin region extending from near Bateman's Bay in the south and extending north to include the Newcastle area and Nelson Bay and west to Lithgow.

This Sydney Basin area involves a GIS-based raster grid of more than 200 million pixels and covers an area of 131,700km2.

Figure 1. Preliminary Landslide Inventory and Susceptibility Zoning for the Sydney Basion region including the Wollongong - Sydney - Newcastle urban areas.

Figure 2. Selected enlarged areas of the Sydney Basin Landslide Susceptibility Model.

The regional Illawarra Landslide Inventory, complemented by additional coverage of the Sydney and Central Coast region provided by the Geoscience Australia's National Landslide database (downloaded from the GA online web portal), has been used to develop the 'training data set'. This combined inventory includes 575 'slide' and 'flow' category landslides within the geological extent of the Sydney Basin. This training data, and hence the output model, will be greatly improved as the Landslide Inventory coverage of this area is enhanced and we are working towards that.

Figure 3. SE Australian area of the combined Landslide Inventory for Australia. Incorporates the UoW Landslide Inventory, Miner and Dahlhaus's Inventory for South Western Victoria, Mineral Resources Tasmania's Landslide Inventory of Tasmania and Geoscience Australia's National Landslide database currently includes over 5,000 landslides.

Over the next 1 - 2 years, as the GIS-based data sets are assembled and further refined, with increased resolution as well, and as the Landslide Inventory coverage is developed and refined with external assistance, we are confident that this wider area regional modelling will be successful. This type of regional modelling, which can be carried out at quite high resolution (large scale) given input data of sufficient resolution, will assist decision makers determine which areas need further landslide zoning investigations, as are also recommended by the guidelines published in AGS 2007.