1. Introduction

Section:

ChooseTop of pageAbstract1. Introduction <<2. Susceptibility assessm...3. Hazard assessment4. Hazard ranking5. Discussion6. Conclusion

The widely reported Scottish debris flow events of August 2004 were caused by rainfall substantially in excess of the norm, with some areas experiencing over three times the 30-year monthly average, and storm intensities of up to around 150mm/h (Winter et al., 2005, 2006, 2009a, 2010).

Long-lasting and intense rainfall led to a large number of landslides, in the form of debris flows, in the hills of Scotland. Critically, some of these affected important parts of the major road network (Figure 1) linking not only towns and cities but also smaller, remote communities. Notable events occurred at the A83 between Glen Kinglas and to the north of Cairndow (9 August), the A9 to the north of Dunkeld (11 August), and the A85 at Glen Ogle (18 August) (Figure 2).

Figure 1. Map of trunk (strategic) road network in Scotland. Locations of the three main groups of events that occurred in August 2004 are shown

Figure 2. Debris flow at A85 Glen Ogle on 18 August 2004

Although there were no major injuries, 57 people were taken to safety by helicopter after being trapped between the two main debris flows on the A85 in Glen Ogle. However, the real impacts were social and economic, particularly the severance of access to and from relatively remote communities. The A83, carrying up to 5000 vehicles per day (all vehicles two-way, 24 h annual average daily traffic, AADT), was closed for slightly in excess of a day; the A9 (carrying 13 500 vehicles per day) was closed for 2 days prior to reopening, initially with single-lane working under convoy; and the A85 (carrying 5600 vehicles per day) was closed for 4 days. The traffic flow figures are for the most highly trafficked month of the year (July or August). Minimum flows occur in either January or February, and are roughly half those of the maxima, reflecting the importance of tourism and related seasonal industries to Scotland's economy. Substantial disruption was thus experienced by local and tourist traffic, and by goods vehicles.

Debris flows occur with some frequency in Scotland, but in the past they have affected major communication networks only rarely. However, when they do affect roads, the degree of damage to the infrastructure and the loss of utility to road users can have a major economic and social impact. Additionally, there is a significant potential for such events to cause serious injury and even loss of life, although to date there have been no such instances.

Tourism accounts for around 10% of Scotland's GVA (gross value added) (source: www.scottish-enterprise.com), and so the impacts of such events can be particularly serious during the summer months; debris flows have historically usually occurred in July and August as well as in the period October to January. (The GVA figure of 10% may well be significantly higher in the north and west of Scotland, where debris flows most commonly occur.) Notwithstanding this, the impacts of any debris flow event occurring during the winter months, during which period debris flows usually occur between October/November and January, should not be underestimated. The debris flow events of 2004 created a high level of awareness in the media of the effects of landslide activity, in addition to being seen as a key issue by politicians at both the local and national level.

The need to acknowledge such natural processes and act accordingly was recognised by Transport Scotland, an agency of the Scottish government, which is responsible for the trunk road network. Consequently it commissioned a study with the overall purpose of ensuring that Transport Scotland has a systematic assessment and ranking of the hazards posed by debris flows, and has a management and mitigation strategy in place for the Scottish trunk road network. The purpose of the ranking system is to allow the future effects of debris flow events to be appropriately managed and mitigated as budgets permit, thus ensuring that the exposure of road users to the consequences of future debris flows is minimised.

Although the 2004 events provided the impetus for the work that has been undertaken, it is important to note that similar events have occurred in the past in Scotland (Winter et al., 2005) and since the 2004 events (Winter et al., 2009a).

This paper describes the work undertaken to satisfy the first key objective of assessing and ranking debris flow hazards. The methodology used to undertake a pan-Scotland (c. 78 000 km²), geographical information system (GIS)-based, assessment of debris flow hazards (or, more strictly, susceptibility) is described, as is the approach taken to interpret the spatial GIS-based assessment for determining those sections of road alignment subject to debris flow hazards. The hazard scores assigned using this approach were subsequently modified in the light of the results of site-specific inspections.

The ranking of hazards, to give an analogue for risk, based upon the potential exposure of road users to debris flow hazards and the potential socio-economic impacts, is also described. Both a map and a table are used to illustrate the locations of the highest hazard-ranking sites in this paper (see later). At each stage, the broad-based approach was to focus on areas highlighted by the preceding stage. The spatial extent of the susceptibility assessment is comparable to the most extensive such assessment undertaken internationally (e.g. Castellanos Abella and Van Westen, 2007; Dio et al., 2010).

Typically, the rainfall-induced debris flows encountered in Scotland (e.g. Figure 3) are characterised by a small translational sliding failure within saturated, unconsolidated (wet and soft) ground on a hillside. If the sliding material reaches a stream channel, typically an incised gully stream channel, then the failed material may be entrained in the water and accumulate further debris as the gully walls are eroded. Obstacles and areas in which the slope slackens may arrest the flow and encourage the formation of debris dams; after a period of time ranging between minutes and years the dammed material may be reactivated, leading to further erosion, before depositing the material in a fan or debris lobe where the slope slackens at its foot. This will typically be where infrastructure such as a road may be located (Winter et al., 2005, 2009a). This morphological scenario is typical of other regions of the world, including the Republic of Korea (Lee and Winter, 2010), where the overall elevations are similar. In general, open hillside debris flow events do not reach infrastructure located in the lower reaches of valleys in Scotland, and the work described here focuses largely on stream channel-based flows.

Figure 3. Morphology of a typical Scottish debris flow: A83 Cairndow, August 2004

A fuller treatment of the work described herein, which also includes the strategy developed for management and mitigation, is presented in Winter et al. (2005, 2009a).

2. Susceptibility assessment

Section:

ChooseTop of pageAbstract1. Introduction2. Susceptibility assessm... <<3. Hazard assessment4. Hazard ranking5. Discussion6. Conclusion

The susceptibility assessment was carried out within a GIS environment (Harrison et al., 2009), and considered the following main characteristics

Given the regional nature of the assessment, covering all of Scotland with only minor omissions, it was important to use data sets that covered the entire area under consideration, in order to ensure that the assessment was applied consistently over the entire area. It was considered that information regarding each of these could be usefully retrieved from the available data sets, as follows.

This GIS susceptibility assessment was calibrated in conjunction with a working group (which included the authors), making use of their wider experience and knowledge of debris flow events in Scotland, their experience of the road network and, most importantly, the interaction between these two elements. Their collective experience of the investigation and management of debris flows in Scotland was a key element in producing reliable outputs from the first-stage, GIS-based susceptibility assessment, which could then be interpreted to derive hazard and hazard-ranking information, as described later in this paper. The possibility of seismic acceleration as a causal factor was discounted, on the basis that seismic events in Scotland tend to be of a magnitude that produces only low-level accelerations. These would include those that are generally accepted to be less than those that can be felt by humans, and less than those that might be exerted by passing heavy traffic.

2.1 Availability of debris material

For a debris flow to occur there must be an available source of material, usually granular, often with a very wide particle size range and in such a state that it would easily be mobilised by the action of water. Thus the material that has the highest potential for debris flow activity is likely to be non-cohesive. Material that is cohesive, owing to high clay content or intergranular cement, is often more difficult to mobilise. Indeed, Gabet and Mudd (2006) have shown that the potential for debris flow mobilisation is sensitive to particle size distribution, with sandier soils more likely to liquefy than clay-rich soils, owing to their higher hydraulic transmissivity, a factor that was highlighted in the early stages of this work (McMillan et al., 2005).

Lithologies represented by polygons in the British Geological Survey DiGMapGB-50 product were interpreted against a scale that indicated the degree to which the bedrock or superficial unit at the surface would provide non-cohesive granular material as a source of debris. The ROCK_D (BGS Rock Description code) attribute of each polygon was reinterpreted by interrogation in terms of potential mobilisation by water as a debris flow in both the fresh and the weathered/regolith states, allowing a score of between 1 and 10 to be assigned.

In addition, and in the absence of a consistent and reliable data set, an analytical method was developed to determine likely zones of peat accumulation, as such materials are known to be capable of triggering debris flow (e.g. Milne et al., 2009). This was based on the plan and profile curvatures of the ground, and was used to adjust the lithology scores in areas where peat was present (Harrison et al., 2009). It is important to note at this point that the assessment was restricted to debris flows, including those triggered in part by peat slides. The potential for peat slides was not included in the assessment. Both Winter et al. (2009b) and Milne et al. (2009) acknowledge the importance of water-bearing soils, particularly peat (see Section 2.3), located on high, relatively flat ground as trigger materials for gully-constrained debris flows.

Figure 4(a) shows the availability of debris material scoring for Glen Ogle and Inverness, as examples. The diagrams show a higher lithology index for the lower slopes and base of the Glen Ogle valley, where the superficial deposits consist of till and morainic deposits, with some alluvial and River Terrace deposits. The area of high lithology index is more widely distributed in the case of Inverness. In this analysis these materials (which can be seen to be on nearly flat or gently sloping ground) are regarded only as potential source areas for landslide debris, because key contributory factors such as slope gradient are not considered in this stage of the assessment (but see Section 2.5).

Figure 4. Excerpts from GIS showing model results for: (a) lithology; (b) water conditions; (c) vegetation and land cover; (d) stream channels; (e) slope angle for Glen Ogle (left) and Inverness (right), north to top. (Base mapping # Crown Copyright. All rights reserved Scottish Government 100046668, 2013)

2.5 Slope angle

Slope angle self-evidently has a major influence on the balance of stabilising and destabilising forces on all slopes. When the destabilising forces exceed the shear resistance of the materials forming the slope, failure occurs. Therefore the steeper the slope, the greater is the susceptibility of the material to initiate a debris flow.

The slope categories significant in the generation of debris flows that were indicated by Winter et al. (2005) were modified following further experience-based discussions. These were used as the criteria to allocate scores to be included in the overall debris flow hazard assessment in Table 1, with those scores being in the range 0·5 to 10.

Table 1. Criteria used to assess slope angle as part of debris flow hazard assessment

Table 1. Criteria used to assess slope angle as part of debris flow hazard assessment

Score	Slope angle: degrees	Comment
0·5	0–7	Generally stable, and influencing only the run-out characteristics of a debris flow.
1	8–15	Slopes within this range that occurred between a road and an area of debris flow hazard were likely to maintain the movement of the debris flow and facilitate its impact on the road, although it was unlikely to be sufficiently steep to allow the initiation of a debris flow within it.
6	16–30	It appears that debris flows may be initiated on slopes within this range, but it would be equally likely that additional material would be incorporated within this zone.
9	31–45	This slope range is considered the most likely to initiate debris flows, based on the experience of the working group. This would appear to be sensible, in that the peak angle of shearing resistance of dry granular material might be expected to be in this range (BS 8002 (BSI, 1994)).
10	Slope > 45	It is logical that slopes in the >45° class should have a factor or weighting greater than the 31–45° class, in recognition of the increased driving force associated with the increase in the downslope component of shear stress.

Figure 4(e) shows the effect of these classifications in the two example areas. Slope is one of the most significant factors in the initiation of debris flows, and as can be seen in the diagrams, in the low-lying Inverness area the slope index is very low. In Glen Ogle the pale colours indicate high slope values above the A85.

2.6 Combined susceptibility factors

The interpreted data were combined to produce a working model of debris flow, and in the main this entailed assessing, testing and adjusting the scores applied to each facet of the assessment and applying corresponding weightings. The weightings were achieved by using the values in Table 2 and multiplying individual factor grids by a weighting value in order to better reflect the individual importance to the initiation of conditions leading to the triggering of debris flow. This was a detailed, iterative validation process that was undertaken in consultation with, and exploiting the knowledge and experience of, the working group. Different scenarios were modelled interactively for discrete geographical areas in order that real-world examples could be used to validate the model results. The resulting grids for each of the five variables were then added together to arrive at a final value.

Table 2. Weightings for the assessed factors and minimum and maximum values prior to the application of weighting. The factors contained in this table were used to multiply each individual factor before the weighted factors were added together at the end of the process to arrive at a final value for susceptibility

Table 2. Weightings for the assessed factors and minimum and maximum values prior to the application of weighting. The factors contained in this table were used to multiply each individual factor before the weighted factors were added together at the end of the process to arrive at a final value for susceptibility

Factor	Weighting	Maximum value	Minimum value
Lithology	×1	10	1
Water conditions	×1	10	0·1
Vegetation	×0·75	1·2	0·7
Stream channel	×0·75	10	0
Slope angle	×1·25	10	0·1

The results from this assessment are illustrated in Figure 5, over which the trunk road network is overlain, giving a broad indication of where the higher-susceptibility areas are located. Figure 6 shows the results for smaller areas, Glen Ogle and Inverness (as for Figure 4), overlain on the 1:50 000 mapping.

Figure 5. Results of GIS-based susceptibility assessment for Scotland: the areas shaded light to dark grey represent increasingly higher levels of susceptibility (see Figure 6)

Figure 6. Results of GIS-based susceptibility assessment for Glen Ogle (left) and Inverness (right), north to top. (Base mapping # Crown Copyright. All rights reserved Scottish Government 100046668, 2013)

As a result of the very long runtimes of the pan-Scotland GIS data, the project team decided that no account of whether potentially susceptible materials were on, above or below a slope and/or the road asset would be taken. As a result some effectively flat areas of no practical debris flow susceptibility, such as the Stirling Carse, were highlighted in the assessment. These areas were eliminated from further consideration during the hazard assessment phase of the work (see Section 3.1).

3. Hazard assessment

Section:

ChooseTop of pageAbstract1. Introduction2. Susceptibility assessm...3. Hazard assessment <<4. Hazard ranking5. Discussion6. Conclusion

The hazard assessment comprised an interpretation of the GIS-based susceptibility assessment to determine road lengths likely to be affected by debris flow, followed by initially selective site-specific inspections using walkover techniques. The GIS assessment was pan-Scotland, covering the entire road network, whereas the interpretation exercise focused on the trunk road network.

3.1 GIS interpretation

The GIS interpretation was undertaken using the imagery from the GIS-based assessment, national Ordnance Survey mapping at 1:50 000 scale, and low-resolution aerial photography. The interpretative process was thus focused on the morphology of the ground between areas of potential susceptibility and roads. Among other factors, slope angles, the presence of stream channels that might aid the passage of debris, and any potential barriers to flow were considered. Consequently, the interpretation may be summarised as a semi-quantitative/qualitative determination of potential debris flow tracks and run-out zones to determine whether they intersect with the trunk road asset.

The purpose of the interpretative work was to identify lengths of road, or sites, that might be subject to debris flow hazards, as opposed to discrete locations potentially affected by a single stream channel, for example. Thus the interpretation was focused on areas and zones of susceptibility. As an example, Glen Ogle was identified, and the length of high susceptibility and hazard was around 5·5 km in this particular instance. The individual (small) areas of susceptibility (such as those illustrated in the Glen Ogle imagery in Figure 6) were thus aggregated. The interpretative work was carried out on the trunk (strategic) road network by a single team in order to ensure consistency of approach, and was split into two stages.

First, a coarse sift was undertaken, examining the entire network in order to categorise sites. Table 3 highlights the approximately 1700 km length of the network for which the susceptibility assessment indicated that one or other of three categories was appropriate; the remaining 1500 km length of route, not highlighted by the susceptibility assessment, is not included in the table. The three categories used were

Table 3. Results of the first phase of interpretation of the GIS assessment. The trunk road network is approximately 3200 km in length

Table 3. Results of the first phase of interpretation of the GIS assessment. The trunk road network is approximately 3200 km in length

	Route lengths assessed: km	Percentage of route lengths assessed: %	Percentage of network: %
Other (None)	619	37	19
Opportunistic	458	27	14
Main Study	607	36	19
Total	1684	100	53

	(a) those sites for which there was a hazard potential, and further consideration was therefore required (Main Study)
	(b) those for which no further action was required (Other (None))
	(c) those for which action would only be required if major upgrade works, typically a realignment of the road within the existing corridor, were planned (Opportunistic).

The sites for which no further action was required (‘Other (None)') were largely flat areas of land highlighted in the susceptibility assessment (see Section 2.6), or where potential debris flows would propagate in a direction away from the road (without the likelihood of undercutting), and were not taken forward for further consideration.

Second, the Main Study sites were separated into four main categories in order to assign a hazard score, and also to indicate those for which further site-specific assessment (see Section 3.2) was most pressing. Two atypical sections of network (Glencoe and south-east Skye) were highlighted for separate assessment; they are characterised by steep talus slopes, and have been intensively studied in the past. In particular, there were several studies that followed talus slope failures in Glencoe during the late 1980s and early 1990s, but the material is largely unpublished and relatively difficult to access; more generic work on Scottish talus slopes is relatively easily available (e.g. Ballantyne and Harris, 1994). These facts mean that these sites lend themselves to a much more focused and targeted approach being taken as part of the follow-up to this study; as an interim measure these sections were initially scored as for Main Study sections. The results of this second phase are summarised in Table 4.

Table 4. Results of the second phase of interpretation of the GIS assessment

Table 4. Results of the second phase of interpretation of the GIS assessment

	Route lengths assessed: km	Percentage of main study route lengths: %	Percentage of route lengths assessed: % (see Table 3)	Percentage of network: %	Initial hazard score
Priority 1	135	22	8	4	80
Priority 2	154	25	9	5	60
Priority 3	160	26	9	5	40
Priority 4	112	19	7	4	20
Separate assessment	46	8	3	1	80
Total	607	100	36	19	–

Typical results are illustrated in Figure 7. These are stored and viewable within a GIS environment; in Figure 7 they are overlain on the 1:50 000 mapping, and lengths designated as Priority 1 to 3 are shown.

Figure 7. Sections of route length categorised as Priority 1 (red, central section), Priority 2 (brown, southerly section) and Priority 3 (orange, northerly section). Digital Ordnance Survey imagery at 1:50 000 is shown, but the image itself is not to scale, north to top. (Base mapping # Crown Copyright. All rights reserved Scottish Government 100046668, 2013)

One of the constraints of the work was the lack of an exhaustive inventory of past events upon which the temporal aspect of the hazard could be predicated. Such aspects were accounted for by using the knowledge of past events of those undertaking the work (primarily the first and third authors), and also by emphasising the importance of evidence of recent and historical instability during the site-specific inspections; where such evidence was found, the hazard scores were increased (see Section 3.2).

3.2 Site-specific studies

The site-specific studies were intended to validate the hazards derived from the interpretation of the GIS-based assessment, to provide an interpretation of data not available during that process, and to provide an interpretation at a more detailed scale than could otherwise be provided. The work for each site was divided into three stages, as follows.

	(a) Desk study. These activities were intended to be carried out prior to embarking upon site-based activities, and included familiarisation with the location from national mapping and the results of the GIS-based assessment. In addition, newly available information in the form of high-resolution aerial photography was incorporated into the assessment.
	(b) Preliminary site inspection. This was intended to allow a provisional, but necessarily limited, view of the site setting. This was achieved by a drive-through of the length of road in question, with the inspector as a passenger, stopping as necessary to observe and note features from road level.
	(c) Detailed site inspection. This process essentially completed the hazard assessment process by relating the information considered thus far (which was either image/data-based or a physical view from a remote location) to the ground itself. In practice, the detailed site inspection comprised a walkover from road level and excursions up slope (or down where necessary) as required, but typically every 0·5 km to 1·0 km.

This stage of the work was primarily intended to adjust the scores already obtained, where appropriate. At each of the above stages, hazard scores were assigned in the categories of water, instability, slope/topography, and vegetation and land use. Photographs were taken to illustrate features and the decisions made to adjust scores. These scores were added to or subtracted from those already obtained from the interpretation of the GIS-based assessment (see Section 3.1), and thus built on the work already carried out. Factors such as deforestation and local evidence of instability increased the hazard score, whereas the existence of large culverts beneath the road (potentially allowing the debris to pass through) decreased it. In this way a greater degree of detail, not available within the national data sets used in the GIS-based susceptibility assessment, could be accounted for.

For example, the influence of vegetation on slope instability is complex, and although the simple approach to vegetation applied at the susceptibility stage (Section 2) was considered appropriate, a greater degree of rigour could be applied during the still rather broadly based site-specific inspection stage. It is recognised that negative effects on stability may result from works related to planting for commercial forestry (e.g. ploughing and ditching) and animal grazing (Winter and Corby, 2012), not least as these activities have the potential to redirect water into undesirable areas. Additionally, wind throw, when trees become over-mature, can be a significant hazard in its own right, in addition to encouraging slope instability when the ground is opened up to water ingress as root structures are torn from the ground. Even where vegetation is clearly a positive influence, the challenge remains one of quantifying the effects adequately for design purposes (e.g. Greenwood et al., 2004). Such issues are considered to be best dealt with at a site level when specific action in terms of mitigation (Section 5) is planned. A more comprehensive treatment of the effects of vegetation on slope instability, including root reinforcement and other factors, is given by Coppin and Richards (2007).

Further, the magnitude of the change in scores was limited in order that the work at this stage did not reflect disproportionately on the hazard assessment. This allowed the potential to highlight, for further and more detailed consideration, specific sites for which the scores might change radically as a result of this process. In the event, large changes were not generally manifest. However, the scores for a significant number of the sites were adjusted upwards in the light of evidence of recent and/or historical instability, among other factors.

The selection of sites for more detailed site-specific work was based on the priorities set out in Table 4, but also took account of the initially limited availability of high-resolution aerial photography, the value of which, in terms of understanding the setting and the history of activity in an area, is illustrated in Figure 8. In addition, there were insufficient resources available to undertake all of the site-specific work planned in the first year. The Priority 1 and 2 sites for which aerial photography was available at the time broadly coincided with the budget available for site inspections, and these studies were undertaken during 2007.

Figure 8. Aerial photographs showing A85 Glen Ogle debris flows of August 2004 and other key features (north to top): (left) 4 km (vertical) by 3 km (horizontal), National Grid reference (NGR) at south-west corner NN 557 250, or 2557 7250; (right) 1 km by 1 km, NGR at south-west corner NN 570 260, or 2570 7260. (1) Northerly debris flow: (a) potential source areas; (b) debris track; (c) runout/debris fan; (d) subsequent carriageway repair. (2) Southerly debris flow: (a) potential source areas; (b) debris track; (c) runout/debris fan; (d) subsequent carriageway repair. (3) Historic rock falls. (4) Other debris flows assumed to have occurred in August 2004. (Licensed to Transport Scotland for PGA, through Next Perspectives^TM)

Typically, the results of the site-specific inspections resulted in increased hazard scores. This meant that those sites that had been inspected generally had higher average hazard scores than those that had not been inspected. Therefore, in order to ensure that the management and mitigation actions could proceed in a timely fashion and on a rational basis, without bias towards the inspected sites, the scores for sites not inspected were adjusted by applying the average score from those sites that had been inspected (Winter et al., 2009a). Further batches of sites have since been studied during 2008 and 2009, with the result that only a very small number remain outstanding (owing to the remaining unavailability of high-resolution aerial photography). Further study of the two sites identified for separate assessment is currently being undertaken.

4. Hazard ranking

Section:

ChooseTop of pageAbstract1. Introduction2. Susceptibility assessm...3. Hazard assessment4. Hazard ranking <<5. Discussion6. Conclusion

The risk to life and limb to which road users may be subject from debris flow hazards was identified at the outset of this project by Transport Scotland's senior management as the primary concern, with socio-economic impacts being secondary (but nonetheless important). In this work the Cruden and Varnes (1996) definition of risk was used, as

$R=HEV$

where R is the risk, H is the hazard, E denotes the elements at risk, and V is the vulnerability of the elements at risk to the hazard.

The result of this equation is herein described as hazard ranking, R_H, as it is recognised that the work reported does not consider all aspects of risk, such as, for example, a full quantification of social and economic risks. Notwithstanding this, some might refer to the hazard ranking as a form of ‘relative risk'.

The hazard was determined as described in Section 3. The elements at risk, namely the road and the associated road users, are either present or not at a given plan location, and therefore correspond to unity where a road is present and zero where a road is not present. The vulnerability equates to risk to life and limb of road users and, to some extent, the socio-economic impacts, including diversionary effects, of temporary closures due to landslides. This allows a simplification of Equation 1, and for the purposes of this study it may be rewritten as

$R_{\text{H}}=H{E}_{\text{X}}$

where E_X represents the combined binary elements at risk, and the vulnerability of road users to life and limb risks and the potential socio-economic impacts, referred to as ‘exposure' herein.

Traffic levels effectively represent life-and-limb exposure (potential for injury or death). Although sightlines and other factors that influence visibility of the road ahead could be used to refine the exposure of life and limb (e.g. McMillan and Matheson, 1997), to a large extent traffic flows relate to the type of road alignment in place, and thus in broad terms to typical sightline lengths. The considerable route lengths under consideration, and the regional nature of the study, meant that an approach that did not incorporate sightlines and related factors directly was considered appropriate. Similarly, the socio-economic aspects of exposure may be represented not only by the traffic flow, in terms of severance and amount of traffic delayed, but also by the existence, length and quality of any diversion necessary if the road were to be blocked.

As with the different elements that make up the hazard assessment, so the different elements of exposure must also be added together in order to achieve an overall score. Relevant categories were determined, and scores were then assigned for both traffic flow and diversionary aspects of exposure for each site. The scores for these individual factors were weighted to reflect their relative importance, and then summed to produce the overall exposure score.

The traffic categories used by Transport Scotland reflect the traffic flows over the entire network. The lowest flow category comprises those roads with an AADT of less than 10 000 vehicles per day. It was apparent at the outset that this lowest category would cover a large proportion of the hazardous sites identified, and would not allow effective differentiation between them. A decision was therefore made to use alternative categories. These new categories and their associated exposure scores (E_XT) were defined as follows.

The categories were determined by inspection of the available traffic flow data, with the intention that they should be chosen to ensure that routes of different character were clearly separated.

The diversion scores (E_XD) were based upon an evaluation of the potential consequences of a closure at each site. Where the diversion was short and effective (e.g. by other trunk and/or A-road), the consequences were defined as ‘Limited'. Where the diversion was long, by difficult means (e.g. C, D and/or unclassified road), or did not exist (in practical terms), the consequences were defined as ‘More significant'. ‘Significant' represents the middle ground between these two extremes, and the diversion scores were defined as follows.

Clearly, the judgements as to whether the consequences of diversions to avoid event locations are subjective and, to some extent, depend on journey origin and destination. These judgements were, however, founded upon extensive experience of the operation, maintenance and management of the road network.

For any given site, weightings were then applied to the two exposure scores. The two weighted scores were then added together to give a total score for exposure. The weightings applied reflect the paramount importance of reducing the exposure to risks related to life and limb of the travelling public, and for this reason the traffic score was weighted more heavily than the largely disruption-focused diversion score. The use of the weightings potentially allows rapid experimentation with different balances between the traffic and diversionary scores. It should, however, be noted that the traffic score does itself include significant elements that relate to the potential disruption to road users.

The final exposure score is thus given by

$E_{\text{X}}=\left(E_{\text{XT}}\times 1.0\right)+\left(E_{\text{XD}}\times 0.5\right)$

Accordingly, Equation 2 may be rewritten as

$R_{\text{H}}=H\left[\left(E_{\text{XT}}\times 1.0\right)+\left(E_{\text{XD}}\times 0.5\right)\right]$

It was thus possible to compute the overall hazard-ranking scores. The final hazard-ranking scores, and the process by which they were obtained, are set out in much greater detail than is possible herein by Winter et al. (2009a). In discussion with the end user of the work (Transport Scotland), it was decided that those sites with the highest hazard rankings (a score of 100 or more; Figure 9) should be subject to management and/or mitigation activities. Table 5 shows the results for the assessments following the full site-specific inspection programme for 2007 to 2009, which comprises 61 sites.

Figure 9. The 66 sites with a hazard ranking score of 100 or greater, north to top. (Base mapping # Crown Copyright. All rights reserved Scottish Government 100046668, 2011)

Table 5. Sites with a hazard ranking score of 100 or greater, following the 2007–2009 programme of site-specific inspections

5. Discussion

Section:

ChooseTop of pageAbstract1. Introduction2. Susceptibility assessm...3. Hazard assessment4. Hazard ranking5. Discussion <<6. Conclusion

Events that have taken place since the prioritisation of hazards (Table 4) and the ranking of hazards were first set out provide a useful qualitative measure of the success of the approach taken to hazard assessment and ranking. Although events in August 2006 in the Helmsdale area did not generally reach the A9 road, they were coincident with the routes coded as A9-11 and A9-12 in Table 5. Similarly, events at the A83 Rest and be Thankful, particularly those in 2007, 2008, 2009, 2011 and 2012, coincide with the route coded as A83-02.

Other events that pre-date the assessment, such as earlier events at the Rest and be Thankful, and those at Loch Lochy (A82-17), Cairndow (A83-05) and Loch Shira (A83-07), all provide examples of the assessment of both hazard and their hazard ranking fitting into the overall pattern of movements adjacent to the Scottish trunk road network. Other events that relate to the local road network in the Highland Council region of north-west Scotland are reported by Winter et al. (2009b), and correlate well with the pan-Scotland susceptibility assessment.

The foregoing observations lend confidence to the ongoing actions that have been and are being implemented on the basis of the hazard assessments and rankings. These actions have been divided into management actions for high hazard-ranking sites, involving exposure reduction, and mitigation for very high hazard-ranking sites, involving hazard reduction (Winter et al., 2011).

Management of hazards and their attendant risk has formed the primary action. Such actions are generally lower cost, and include educational activities, permanent geographical warnings, and temporally responsive actions (Winter, 2013).

Educational activities undertaken include the production of a leaflet that sets landslides in the broader context, as well as providing advice for road users. This approach is in the process of being extended to fixed information boards that are currently being developed in collaboration with Scottish Natural Heritage, the Scottish National Park authorities, and Forestry Commission Scotland.

Permanent geographical warning signs are also used to indicate the presence of a hazard, and have been used to indicate the start and end points of those sections identified in Table 5. These are the red triangles on white background with the black landslides logo that are familiar in so many regions of the world.

Temporal warnings through responses to events and to rainfall conditions that lead to an increased likelihood of event occurrence form a critical part of the exposure reduction activities. These are managed through the process of detection, notification and action (DNA) (Winter et al., 2005)

Specific actions include media announcements related to event occurrence and severe rainfall warnings, and also the provision of landslide patrols in marked vehicles, road closures, and traffic diversions to avoid high-risk areas. At the A83 Rest and be Thankful there has been a trial of landslide warning signs, with the addition of flashing warning lights that are operated during periods of severe rainfall and otherwise higher risk (Winter et al., 2011). Progress towards a proactive system of landslide warnings has been described by Winter et al. (2010), and these actions are ongoing.

To date, mitigation, or hazard reduction, measures have been largely limited to the site at the A83 Rest and be Thankful. The main activity has been the installation of debris catch fences and works to improve the drainage regime in the immediate vicinity of the road.

The challenge with hazard reduction is to identify locations of sufficiently high hazard ranking to warrant spending significant sums of money on engineering works. The view that the costs associated with installing remedial works over long lengths of road are both unaffordable and unjustifiable has, to some extent, prevailed. Moreover, the environmental impact of such engineering works should not be underestimated, having a lasting visual impact at the least and potentially more serious impacts. It is considered that such works should be limited to locations where their worth can be proven, and where they also fit within the priorities of the broader trunk road improvement programme. Work on the quantitative risk assessment of higher hazard-ranking sites is under way, and will help to support the decision-making process for the provision of engineering works in the future. This approach seems appropriate compared with other regions of the world where, in general, only high levels of risk to life and limb attract significant and extensive expenditure on engineered works (e.g. Anon, 2005; Hong et al., 2005; Versace, 2007; White et al., 2007; Winter and Bromhead, 2012).

Notwithstanding the foregoing, it is important that maintenance and construction projects take the opportunity to limit any hazards at the design stage by incorporating, where suitable, measures such as higher-capacity or better forms of drainage, or debris traps. In particular, critical review of the alignment of culverts and other conduits close to the road should be carried out as part of any planned maintenance or construction activities.

There are three broad approaches to hazard reduction in the context of the effects of debris flow on roads: works involving road protection, prevention of debris flow, and road realignment, and these are explored in more detail by Winter (2013).

In this case of road protection it is accepted that debris flows will occur, and provision is made to protect the road by the use of means such as debris basins, lined debris channels to move material downslope where potential storage areas on the hillside are limited, debris flow shelters, barriers and fences. As part of work at the A83 Rest and be Thankful site, defined as one of the higher-risk sites on the Scottish trunk road network, fully flexible barriers have been installed (Figure 10). Rigid barriers such as check dams and baffles may slow and partially arrest flows within a defined channel, and on hillsides may protect larger areas where open hillside flows are a hazard.

Figure 10. Flexible debris flow fence at the A83 Rest and be Thankful in Scotland

The engineering solutions applicable to the prevention of debris flow depend greatly upon individual circumstances. Small debris-flow-triggering events can emanate from relatively large source areas, and be initiated high on the hillside above the road. There may be particular conditions where conventional remedial works and/or a combination of techniques such as gravity retaining structures, anchoring or soil nailing may be appropriate. However, in general terms, the cases where these are both practicable and economically viable are likely to be limited. The link between debris flows and intense rainfall has been established previously in this paper. As a result, effective runoff management can reduce the potential for debris flow initiation. In the circumstances of the debris flows that occurred in the summer of 2004, it is considered that on-hill drainage improvement would have had little impact, because of the scale of the events. In other locations and situations, positive action to improve drainage might well have a beneficial effect. Such measures could include improving channel flow and forming drainage around the crest of certain slopes to take water away in a controlled manner.

Road realignment may also be undertaken, but it is likely that it will form part of broader route improvement activities in order to improve the overall service in terms of alignment and junction layout, and in particular to reduce accidents and achieve compliance with current design standards. In cases where the debris flow hazard ranking is high, and other factors indicate that some degree of reconstruction is required, road realignment may be a viable option. Indeed, a study to examine the options for a section of the A83 route, including realignment, has recently been announced by Transport Scotland, with the debris flow hazard and risk at the Rest and be Thankful forming a significant part of the focus for the planned work.

6. Conclusion

Section:

ChooseTop of pageAbstract1. Introduction2. Susceptibility assessm...3. Hazard assessment4. Hazard ranking5. Discussion6. Conclusion <<

This paper reports on the approach taken to a regional study of hazards and risks from debris flow as they affect the Scottish road network.

The initial susceptibility assessment was carried out in a GIS environment, and used well-established data sets for which full national coverage was available. The process of translating susceptibility into hazard required a much more hands-on approach in order to determine the lengths of road that might potentially be reached by susceptible materials that have been mobilised. In addition, and in order to define the hazard adequately, the use of on-the-ground site assessments was deemed essential, as was the use of high-resolution aerial photography, which was not available during the earlier stages of the study. Both activities proved their worth by allowing the identification of features that were not evident during the earlier phases of the work.

The primary vulnerability was considered to be the exposure of road users to life-and-limb risks, although social and economic issues were also considered. These issues were captured by means of traffic flow data and an assessment of the potential for traffic diversion.

All elements of the work were carried out with close collaboration between engineering geologists, geotechnical engineers and infrastructure engineers, each with extensive experience both of the trunk road network and of debris flow occurrences that have adversely affected the network in the past. This approach proved its worth throughout each phase of the work. It seems clear that events that occurred during the currency of the work and after its completion improved confidence in the use of the methodology to prioritise management and mitigation actions within the available budget. The main focus of such actions has been management to reduce the exposure of road users to the hazards, with mitigation actions to reduce the hazard being carefully targeted at those sites associated with the highest risks.