CN River Attack Track Risk Assessment System (RATRAS)

CN River Attack Track Risk Assessment System (RATRAS) Michael Porter, Dr. Iain Bruce & Mark Pritchard BGC Engineering Inc., Suite 500, 1045 Howe Street, Vancouver, British Columbia, V6Z 2A9, Canada

Tim Keegan and Brian Abbott Canadian National Railway, Engineering & Environmental, 16th Floor, 10004-104th Avenue, Edmonton, Alberta, T5J 0K2, Canada

4,753 Words

CN River Attack Track Risk Assessment System (RATRAS)

Michael Porter, Tim Keegan, Brian Abbott, Dr. Iain Bruce and Mark Pritchard

CN’s rail network in British Columbia parallels major river systems. Consequently, River Erosion (RE) and River Erosion-Earth Slide (RE-ESl) ground hazards present a significant risk for service disruptions and derailments. A review of loss records indicates that (RE) and (RE-ESl) hazards have historically impacted CN’s rail operations in British Columbia with a frequency of approximately 0.005 incidents per mile per year. In response, CN has collaborated with BGC Engineering Inc. to develop a quantitative risk management tool entitled the “River Attack Track Risk Assessment System” (RATRAS). Once implemented, the system will facilitate a rational prioritization of monitoring and mitigation activities.

Key Words: River Erosion, Landslides, Risk Management

INTRODUCTION CN operates roughly 2,700 miles of mainline track in the Province of British Columbia, including some 1,455 miles of the former BC Rail network (Figure 1). The province comprises rugged, mountainous terrain that is subject to considerable seasonal variability in temperature and precipitation. Due to recent glaciations, the river systems in BC are in an immature state of their development and are relatively geomorphically active.

Conditions dictate that railways and

other linear transportation and utility infrastructure are typically located along the valley bottoms of major river systems. This paper describes the development of a quantitative risk assessment procedure to manage the railway’s exposure to river erosion hazards.

RAILWAY GROUND HAZARD MANAGEMENT Significance The terrain, climatic and geomorphic conditions described above expose the railway to hydraulic erosion (including river erosion), landslides, subsidence, and snow avalanches. Collectively, these are termed “railway ground hazards”. Although train derailments related to ground hazards are relatively infrequent, their consequences are significant.

Over the past 10 years,

ground hazards account for only 5% of CN’s derailments in Western Canada, but have resulted in 20% of derailment-related economic losses (1). These trends are mirrored in the pipeline industry (2).

Classification of Railway Ground Hazards Effective management of risk requires that hazard and risk scenarios be identified and characterized.

CN employs a system developed by Keegan (1) to classify railway ground

hazards leading to a loss. As illustrated in Figure 2, the system accounts for two main factors: a)

material type (e.g. water, soil and rock, and snow and ice), and

b)

process (e.g. hydraulic erosion, slides, falls, avalanches, etc.)

Complex ground hazards comprise those involving more than one process. The classification system refers to river erosion as Level III: Sub-aqueous flow erosion, and Level IV: Channelized flow erosion. Figure 3 describes the Level V classification of channelized flow erosion, otherwise described as river erosion.

Review of CN’s loss records has demonstrated that River Erosion (RE) and the compound mechanism River Erosion – Earth Slide (RE-ESl) are among the leading mechanisms of loss. Keegan et al. (3) have previously described the (RE-ESl) hazard. Figure 4 illustrates a typical incident that occurred along the Skeena River in February 2002.

Both CN and the former BC Rail have maintained records of River Erosion (RE) and River Erosion-Earth Slide (RE-ESl) incidents impacting their respective rail operations. Of the 208 incidents tabulated in Table 1, only 5 (roughly 2.5%) resulted in train derailments. The frequency of derailments is probably constrained by extra precautions typically undertaken by the railway during periods of flooding (e.g. special patrols) and by the fact that earth slides often develop over periods of hours to days, facilitating detection prior to derailment.

The data suggest that CN has been more frequently impacted by the compound failure mechanism of river erosion-earth slide (RE-ESl) (0.0061/mile*year) than by pure river erosion (RE) incidents (0.0009/mile*year). Conversely, BC Rail data shows more river erosion (RE) incidents (0.0037/mile*year) than river erosion-earth slides (RE-ESl) (0.0003/mile*year). The reason for this difference is not clear but may stem from incident reporting mechanisms, the terrain, or some combination of these factors.

DEVELOPMENT OF THE RATRAS In response to the Skeena River incident and others, the need was identified for a rational system to: a)

identify and characterize hazard sites,

b)

subjectively estimate associated risks, and

c)

effectively allocate inspection and mitigation resources.

Consequently, CN and BGC Engineering Inc. (BGC) have collaborated in the development of the “River Attack Track Risk Assessment System” (RATRAS). RATRAS provides a framework for managing risks associated with a range of river erosion hazards, as well as river erosion – earth slide hazards. Ratings are based upon sound engineering and geoscience practice.

System Objectives and Requirements The overall objective of the RATRAS is to proactively identify areas subject to hydraulic erosion hazards before they lead to service disruption or a train derailment, and to establish an objective rationale for prioritization and early intervention. It is intended that the system facilitate the full cycle of risk management, including: a)

identification and classification of hazards,

b)

quantitative risk estimation,

c)

inspection protocols,

d)

prioritization and implementation of remedial work,

e)

documentation of work and re-rating, and

f)

communication with stakeholders.

System Design Principles Principles governing the design of the RATRAS comprised: a)

Compatibility - The system had to be compatible with the Canadian Standards document, “CAN/CSA-Q850-97 Risk Management: Guideline for Decision-Makers,” and consistent with the Rockfall Hazard and Risk Assessment (RHRA) methodology that has already been implemented by CN to manage rock fall hazards (4 and 5).

b)

Objectivity - The system had to generate quantitative estimates of the likelihood and consequence of derailment.

c)

Simplicity - The system had to strike an appropriate balance between the level of complexity required for rigorous rating and the simplicity necessary to facilitate cost-effective implementation of the system.

d)

Repeatability - Procedures had to be easily understood and repeatable. It is essential that rating, when guided by appropriate reference material (e.g. illustrated rating forms and field manual) and undertaken by qualified engineers or geoscientists, produce similar scores.

e)

Facility of Data Management - Finally, the system had to lend itself to a data management system that could: •

provide a corporate record of ratings, inspections, incidents, and mitigation work, and

•

facilitate effective communication between CN personnel, consultants, and other stakeholders at each phase of the risk management cycle.

Hazards Considered River Erosion (RE) and the compound mechanism River Erosion – Earth Slide (RE-ESl) involve a number of processes (or elemental failure modes) that could render the track inoperable or lead to train derailment. As outlined by Keegan (1), these include:

a)

Channel Aggradation

Channel aggradation refers to raising the level of streambed when sediment supply exceeds sediment transport capacity. It can lead to burial of a bridge, increased loading on a bridge especially during flooding, erosion due to channel widening, increased likelihood of flooding, debris blockage and bridge overtopping. Aggradation is also a leading cause of stream avulsion. Evidence of aggradation is incorporated in the assessment of the likelihood of the other river erosion hazard types. b)

Channel Degradation

Channel degradation is the general lowering of the channel over a reach of a river or stream. It is often in response to a decrease in sediment supply, the down-cutting of an immature river system or down-cutting into landslide material. Degradation can cause bridge failure by undermining pier or abutment foundations, and can cause earth slides by oversteepening slopes. c)

Local Scour

Local scour involves localized deepening of the channel by erosion caused by vortexes created by obstructions, increased velocity and downward spiraling currents on the outside of a meander or differentially erodible material on the channel bottom. Obstructions can be manmade such as bridge piers or abutments or natural such as bedrock knobs, boulders, gravel bars, or log jams. d)

General scour

This involves localized lowering across a channel due to reduction in the effective width of the channel. Constrictions can be manmade, such as rock berms or bridge approach fills, piers and abutments, or naturally occurring as in the case of alluvial fans, colluvial fans or landslides that encroach and reduce the effective channel width. At the present time, RATRAS considers the combined effects of potential local and general scour.

e)

Ice or log jams

Ice or log jams can cause a localized reduction in the effective depth of the channel. Flooding and avulsion can result, and damage can also occur to bridges either from impact of the ice or debris, or from excessive forces on the bridge caused by the impeded flow. f)

Bank Erosion

Bank erosion is defined as localized loss of the bank material below the water line at stream crossings.

It occurs commonly during high water on unprotected and erodible riverbanks.

Poorly consolidated silts, sands and gravels erode quicker than bedrock, cobbles, boulders or cohesive material. Locations on the outside bend of a meander are more susceptible to bank erosion, as velocities are usually greater and flow direction spirals downward. The most common effect is a reduction of shoulder width and support for the track structure. g)

Encroachment

Encroachment involves the same physical processes as bank erosion, but involves locations where the track runs parallel to a stream valley. h)

Avulsion

Avulsion is a sudden abandonment of one stream channel in favour of another. It is common in alluvial fans, the floodplains of large anastamosed or braided streams and in streams with high bed load or poor confinement. It can be total or partial abandonment of one channel for another. Avulsion can lead to erosion or undermining of the rail grade at the location of the newly occupied or primary stream channel. i)

Earth Slide

Earth slides involve a translational or rotational displacement of surficial materials, usually in response to a loss of toe support or a change in groundwater conditions. In this risk assessment, the evaluation of the likelihood and consequence of earth slides focuses on slides with a toe located in a stream channel below the rail grade. Slope movement can result in a loss of support

or deformation of the rail grade leading to track failure. Slope movements may occur weeks, months, or even years after a flood caused erosion at the toe of the slope.

Event Trees for Risk Scenarios For each hazard type identified above, there is a sequence of events from initiation of the hazard to failure of the rail grade and, potentially, to train derailment. Event trees can be developed to describe each sequence of events.

Figure 5 provides an example tree for

encroachment leading to service disruption and train derailment as a result of either earth slides or grade erosion.

The sequence involves a number of components that need to be assessed: a)

the likelihood of an initiating event (in this case, exceeding the bankfull condition. The “bankfull condition” is defined as the flood stage that occurs, on average, once every 1.5 years (6). Geomorphic work by streams increases dramatically once bankfull conditions are reached or exceeded.),

b)

the likelihood of river erosion (in this case, encroachment) given the bankfull condition is exceeded (PBE);

c)

the likelihood of earth slides, given encroachment (PBE | ENC);

d)

the likelihood of service disruption, given earth slides, defined as the track vulnerability to earth slides (VES);

e)

the likelihood of derailment, given service disruption, calculated using an “avoidance factor” (AF) similar to that used in CN’s RHRA;

f)

the consequence of service disruption and derailment; and,

g)

the likelihood of encroachment leading to grade erosion and service disruption, defined as the track vulnerability to encroachment (VENC).

An assumption implicit in Figure 5 is that the total probability of service disruption from both river erosion and river erosion-earth slide cannot exceed 1.0.

Definition of an initiating event (in this case the bankfull condition) provides an opportunity to conduct real-time river erosion risk monitoring in the future. For example, on gauged streams the probability of exceeding the bankfull condition could be replaced by the ratio of current discharge to bankfull discharge. Risk sites for streams along the rail network that are in flood at any given time could be targeted for increased inspection or other operational protocols such as a temporary reduction in train speed.

Estimating Event Probabilities Each branch of the event tree illustrated in Figure 5 requires an estimate of probability or event likelihood. A number of options are available for estimating event likelihoods as part of quantitative risk assessment. Some of these include: a)

failure statistics;

b)

numerical modelling;

c)

subjective probability; and,

d)

quantitative attribute methods.

The Quantitative Attribute Approach An attribute-based approach was adopted for calculating component event probabilities within the RATRAS following a review of the available options, as discussed below.

Quantitative attribute methods are a means of developing an inventory of conditions that indicate a system is more or less likely to fail, and, using that inventory, to systematically assign a probability of failure within a quantitative risk assessment. One of the best-documented uses

of attribute methods is that by Muhlbauer (7) who developed the approach to improve decisionmaking for pipeline integrity management. BGC uses the approach to rank the potential for pipeline failure resulting from geohazard exposure (2; 8; 9). The BC Ministry of Transportation uses a similar semi-quantitative method to assess the risks from rock fall.

The approach involves quantifying engineering judgement by developing a set of attributes that provide an indication of probability of hazard occurrence and the probability of system failure, should the hazard occur. The possible responses for each attribute (e.g. stream classification, bank materials, and the presence of obstructions) are assigned numerical scores between 0 and 1 that are multiplied to provide an estimate of the probability of hazard occurrence.

Similarly, attributes relating to system vulnerability (such as the distance between the stream and the railway) are assigned scores that are combined to provide an estimate of system vulnerability. The product of the probability of hazard occurrence and system vulnerability gives the estimated probability of failure. Once failure predictions are calibrated using failure statistics and engineering judgement, the results are sufficiently accurate to guide risk management activities such as the allocation of inspection, monitoring, and mitigation resources.

Use of quantitative attribute methods to assign event probabilities within a quantitative risk assessment offers several inherent advantages: a)

they provide an inventory and record of site conditions, and are thus ideal for tracking changes in conditions over time, meeting the requirement for a dynamic rating system;

b)

they provide a more transparent and repeatable rating process since site attributes are easier to characterize in a systematic manner than are event probabilities; and,

c)

the attribute response scores can provide a reasonable and defensible estimate of their influence on the probability of hazard occurrence or system vulnerability once calibrated using failure statistics, numerical modelling, and engineering judgement.

Many attributes affect the likelihood of river erosion hazards and failure of a railway system.

The challenge is to select a minimum, but sufficient number of attributes to describe a system and estimate the likelihood of hazard occurrence and system failure. For each RATRAS hazard, the total number of attributes collected for assessment of likelihood and vulnerability has been limited to 9 or 10.

BGC drew on experience with river erosion risk assessment for linear infrastructure, including pipelines, in selecting the attribute types, responses and collection procedures utilized in the RATRAS. The selection and weighting of attributes for assessing channel instability are also supported by a number of technical references (6; 10; 11; 12).

Attribute data is initially sourced from 1:10,000 scale air photographs and from CN’s “as-built” bridge drawings. Field verification and ongoing inspection is then carried out via helicopter, boat and foot inspections. At some sites, final evaluation of the effectiveness of mitigation will require detailed engineering studies. In these cases, conservative rating values will be used until detailed studies have been completed. Hazard Likelihood Attributes “Hazard Likelihood Attributes” are employed in assessing the probability that a given hazard will occur.

Table 2 provides a summary of the “Hazard Likelihood Attributes” utilized by the

RATRAS. Note that attributes may be common to several, but not necessarily all, hazard types.

Vulnerability Attributes “Vulnerability Attributes” reflect the probability that the railway will suffer an adverse impact should a hazard occur. Table 3 summarizes vulnerability attributes used in the RATRAS. Derailment Avoidance Factor Similar to the RHRA system, the RATRAS employs an Avoidance Factor (AF) to account for the conditional probability of train derailment given that the rail grade becomes inoperable. Table 4 describes the attributes considered in assessing the Avoidance Factor (AF), where: AF = ƒ (Warning Signal, Patrol Frequency, Train Speed).

(1)

Based on the historical data, it appears that a reasonable range for AF is between 0.01 and 0.10. Consequence Factor Derailment consequence (e.g. loss of life, property damage, and environmental impact) depends greatly on the terrain and the geometry below the rail grade. In the RHRA, this is accounted for by applying a Consequence Factor (CF) to each of the four different derailment terrain categories identified in Table 5. Identical Consequence Factors are used by the RATRAS.

Since RATRAS ratings pertain to sections of track located close to streams, most sections fall into consequence categories C and D. Consequence B categories may apply to some sections that are located far from a river, but still subject to river erosion – earth slides.

Quantitative Rating Algorithms Referring to the event tree in Figure 5, the likelihood of occurrence for each river erosion hazard (PRE) is calculated as 0.67 times the product of the hazard likelihood attribute response scores. The value 0.67 is the annual probability of a stream reaching or exceeding its bankfull condition.

For each river erosion hazard, the subsequent likelihood of triggering an earth slide (PES |

RE)

is

calculated as the product of the earth slide hazard likelihood attribute response scores. The subsequent likelihood of service outage due to earth slides (VES) is calculated as the product of the earth slide vulnerability attribute scores.

The resulting probability of river erosion and subsequent earth slides causing service disruption is: PSD | ES = 0.67 x PRE x PES | RE x VES

(2)

As indicated in the event tree in Figure 5, the probability of river erosion directly causing service disruption is taken as: PSD | RE = 0.67 x PRE x (1 – PES | RE) x VRE

(3)

Multiple river erosion hazards will apply at each site. A series summation is used to add up the contribution of each hazard type to the total probability of service disruption. This takes the form: PSD all hazards = 1 – (1 – PSD hazard 1) x (1 – PSD hazard 2) x (1 – PSD hazard n)

(4)

The overall derailment risk rating for a site may be expressed as: RRRATRAS = PSD all hazards * AF * CF

(5)

System Calibration Attribute response score values have initially been established based on review of historical incident data, engineering judgment, and experience with pipeline hazard and risk assessment.

Calibration of these scores will take place continuously through the implementation phase of the project.

CONCLUSIONS CN and BGC are developing a quantitative risk assessment tool to manage railway river erosion (RE) and river erosion-earth slide (RE-ESl) hazards.

Risk assessments will be based on

information readily available from air photographs and field observations. The results will guide the allocation of resources for proactive site inspection, monitoring, risk control assessment and mitigation work, and will document changes in site conditions. The RATRAS represents a significant advancement in the management of this category of railway ground hazards, and is expected to have application on other types of linear facilities.

REFERENCES

1.

Keegan, T. 2004. Railway Ground Hazard Classification System, Unpublished.

2.

Porter, M., Logue, C., Savigny, K.W., Esford, F., and Bruce, I. 2004. Estimating the Influence of Natural Hazards On Pipeline Risk And System Reliability.

Proc.

International Pipeline Conference 2004, Calgary, AB.

3.

Keegan, T., Abbott, B., Cruden, D., Bruce, I., and Pritchard, M., “Railway Ground Hazard Risk Scenario: River Erosion: Earth Slide”, Proceedings of the 3rd Canadian Conference on Geotechnique And Natural Hazards, Edmonton, June 2003.

4.

Abbott, B., Bruce, I., Keegan, T., Oboni, F., Savigny, W., 1998, A Methodology for the Assessment of Rockfall Hazard And Risk along Linear Transportation Corridors, 8th Congress, International Assoc. of Engineering Geology, A Global View From The Pacific Rim, Vancouver British Columbia.

5.

Abbott, B., Bruce, I., Keegan, T., Oboni, F., Savigny, W., 1998, Application of a New Methodology for the Management of Rockfall Risk along a Railway, 8th Congress, International Assoc. of Engineering Geology, A Global View From The Pacific Rim, Vancouver British Columbia.

6.

Rosgen, D. 1996. Applied River Morphology, 2nd Edition. Wildland Hydrology, Pagosa Springs, Colorado.

7.

Muhlbauer, W.K. 2004. Pipeline Risk Management Manual, 3rd Edition. Elsevier Inc. and Gulf Professional Publishing.

8.

Savigny, K.W., Porter, M., Chen, J., Yaremko, E., Reed, M., and Urquhart, G. 2002. Natural Hazard and Risk Management for Pipelines.

Proc. International Pipeline

Conference 2002, Calgary, AB.

9.

Esford, F., Porter, M., Savigny, K.W., Muhlbauer, W.K., and Dunlop, C. 2004. A Risk Assessment Model for Pipelines Exposed to Geohazards. Proc. International Pipeline Conference 2004, Calgary, AB.

10.

Kellerhals, R., Church, M., and Bray, D. 1976. Classification and Analysis of River Processes. Asce Journal of the Hydraulics Division, Vol. 102, No. Hy7, July 1976.

11.

FHWA. 1995. Stream Stability at Highway Structures, 2nd Edition.

12.

Thorne, C., Hey, R., And Newson, M. 1997. Applied Fluvial Geomorphology for River Engineering and Management. John Wiley & Sons.

TABLE 1 - River Erosion and River Erosion–Earth Slide Service Disruption Incidents in BC Data Source

Period of Record

Number of Incidents

CN

12 yrs (1992 – 2004) 18 years

105

N/A

208

Former BC Rail Aggregate

103

Frequency 0.007/ mile*year 0.004/ mile*year 0.005/ mile*year

TABLE 2 - Hazard Likelihood Attributes Attribute

Description

Hazard Activity Observations of current or past (for each haz- activity, similar to the initial rock ard type) fall frequency estimate in the RHRA Angle of Attack Obstructions Bed Materials

Bank Materials

Flow direction relative to the bank (-ve = away from bank; +ve = towards bank) Such as rock knobs, abutments or piers in the floodplain, log jams, beaver dams Based on gradation and genesis

Based on gradation and genesis

Stream Classi- After Rosgen (1996), incorpofication rating planform, gradient, entrenchment Control StrucPrevent degradation, such as tures check dams, lag deposits Landslides or Alluvial Fans U/S or D/S Bank Protection

Changing sediment supply and promoting aggradation, degradation, and avulsion Such as riprap, gabions, anchored logs, groynes

Channel Confinement

Incorporates valley setting, floodplain development, and entrenchment Measure of the channel constriction at bridges, landslides, alluvial fans or other obstructions (influences depth of local & general scour and rate and

Relative Effective Stream Width

Responses Active (last cycle of seasons) Inactive (historical records, old field evidence) None (no records; cause/trigger not present) -90 to +90 deg. Yes No Strong Rock (>R2) Weak Rock (R2) Weak Rock (10% of watershed) Few (affecting R2) Weak Rock ( 4 m above 0m 0 to 1 m 1 to 2 m 2 to 4 m 4 to 10 m 10 to 20 m > 20 m 4m

Attribute

Description

Distance to Top of Subgrade or Abutment

Horizontal distance to crest of riverbank

Elevation at Top of Subgrade or Abutment

Height above bankfull elevation

Subgrade Materials

Material directly below track structure. Usually derived from local materials.

Earth Slide Volume

Volume of active or dormant slides, or expected slide volume should a new slide occur Expected velocity, should a slide occur

Earth Slide Velocity

Responses 50 m 50 m Riprap or material suitable for riprap Shot-rock undersized for riprap Clay Sand and Gravel Sand Silt Very Large (> 100,000 cm. m) Large (10,000 to 100,000 cu. m) Medium (1,000 – 10,000 cu. m) Small (< 1,000 cu. m) Rapid (>1.8m/hr) Moderate (13m/month – 1.8m/hr) Slow (13m/month – 1.6 m/year) Very Slow (10/1 10/1 to 5/1