Societal risk evaluation for landslides: historical synthesis and proposed tools

Abstract

When a life-threatening landslide is recognized, the questions most relevant to community leaders are as follows: is our community safe enough, and if not, how much should we spend on protection? A risk evaluation tool that helps answer these questions, which compares landslide risk faced by a group of people to perceptions of tolerable risk, was first proposed in Hong Kong and is now being used widely in Western Canada within a quantitative risk management framework. After more than a decade of sporadic application in Western Canada, challenges of applying this tool are becoming apparent. For example, its use has resulted in landslide mitigation designs that are unaffordable, and it has failed to persuade funding authorities that proposed risk management solutions are a justifiable allocation of governments’ limited resources. This article suggests that the risk evaluation tool designed in Hong Kong should not be universally applied. We propose modifications believed to be more appropriate for Western Canada, including tolerating landslide risk levels similar to other natural hazard types and emphasizing cost-effectiveness of landslide risk reduction options. Although the proposed tools were developed for the sociopolitical context of Western Canada, we hope the discussions included in this article motivate others to modify these risk evaluation tools for other societies and hazard types and that ultimately this will lead to more rational and consistent decisions that, with time, save lives and resources in landslide-prone regions of the world.

Introduction

Scientists and engineers who study landslides tend to focus on the kit of tools, technologies, methods, and judgment that the profession uses for making assessments. Their efforts are commonly directed towards the mechanics of landslide initiation and motion, informing assessment of where and why. This article focuses on what comes next—the public policy and political decisions related to:

• Is my community safe enough from landslides?

• How much of our limited resources should be invested in landslide protection?

• Would that money be better spent to protect against floods, fires, and earthquakes?

• When can evacuees return home?

• How do we communicate the risks and empower stakeholders to help manage risks?

A risk management framework that informs these decisions is well established (Whitman 1984; Fell 1994; Fell et al. 2005; VanDine 2012; Corominas et al. 2014; Porter et al. 2017; ISO 2018) and recent advances, such as the Sendai Framework for Disaster Risk Reduction (UNDRR 2015), suggest that forward-looking and action-oriented approaches will be more widely adopted and further developed. This article focuses on the “risk evaluation” step of a quantitative risk management framework (Fig. 1). Risk evaluation is the comparison of estimated landslide risks with available resources and perceptions of tolerable risks to decide what actions, if any, are needed. This stage informs the decisions listed above, and it brings engineers and geoscientists outside their comfort zones into the realm of social scientists, planners, and policymakers. These decisions require assessment of human perceptions, beliefs, and values, rather than landslide mechanics.

Fig. 1

Landslide risk management framework (adapted from Fell et al. 2005; VanDine 2012; ISO 2018). Risk evaluation (highlighted in red) is the focus of this article

A tool (Fig. 2) that has been proposed as a measure of one particular society’s values related to landslide risk is the Hong Kong societal risk tolerance criteria for landslides (ERM 1998). These criteria (called “FN criteria”) have great utility (if one assumes them to be true) because they condense dynamic and sometimes irrational human feelings to a simple set of lines and numbers. The risk evaluation tool allows one to plot landslide risk estimated for a group of people (in terms of landslide probability and potential fatalities) and judge whether the people at risk ought to accept the risk outright (broadly acceptable), spare no cost in reducing risk (unacceptable), or tolerate the risk so long as the cost of further risk reduction is “grossly disproportional” to the benefit gained (ALARP, meaning “as low as reasonably practicable”). This simplicity has led to its adoption in other societies beyond Hong Kong.

Fig. 2

Societal landslide risk evaluation tool (FN criteria). This particular example was developed in Hong Kong (ERM 1998) and has been referenced in Canada, although a different version without the “broadly acceptable” zone is more widely used in Hong Kong

FN criteria appear occasionally in landslides literature (e.g., Sobkowicz et al. 1995; Jakob and Weatherly 2005; Friele et al. 2008; Wang et al. 2014; Macciotta et al. 2016; Keller 2017; Macciotta et al. 2019), and in Western Canada, FN criteria for landslide and steep creek risk management are becoming a common reference (Porter and Morgenstern 2013; Porter et al. 2017; Macciotta and Lefsrud 2018). Consultants and municipal governments in Western Canada began referring to FN criteria as early as 1995 (Sobkowicz et al. 1995) and, by 2005, after some public consultation, generally accepted that Canadian perceptions and values related to landslide risk are similar to those of Hong Kong (Tappenden 2014; Porter et al. 2017; Macciotta and Lefsrud 2018). Although there has not been universal acceptance, since 2005, FN criteria have informed land use decisions at more than 50 landslide and steep creek flood hazard areas and, as of 2020, are being widely referenced with discussion of incorporating them into municipal and provincial bylaws, codes, and guidelines.

This article proposes an evolution of societal landslide risk evaluation, based on the authors’ experiences. It is intended to be a resource for risk analysts and community leaders who reference FN criteria when confronted with landslide hazards and the questions listed in the opening paragraph. The scope of the article is specific, but we believe the underlying concepts have much broader application. This article focuses on a subset of potential landslide risk scenarios—that which involves a catastrophic landslide that impacts a group of tens to thousands of people, in which life loss risk controls decision-making and in which a quantitative risk management framework is employed. In Hong Kong, this scenario mostly involves high-rise buildings located at the foot of steep slopes (Malone 2005). In Western Canada, this scenario has tended to involve communities of single-family homes located on alluvial fans that are threatened by debris flows, rock avalanches, and debris floods (Church and Jakob 2020). The tools and methods described here could be applied to other hazard types (e.g., floods, wildfire, dam breach), and many of the considerations described here would apply equally to risk management scenarios that focus on life loss risk to individuals.

This article is structured in three sections. The first section provides background information, including definitions of risk evaluation criteria and terms, and a brief history of FN criteria, highlighting points that appear to be overlooked or misunderstood. The second summarizes Western Canadian case histories where FN criteria have informed landslide risk decisions. The third proposes societal landslide risk evaluation tools. We conclude that the risk evaluation tool developed in Hong Kong may not be directly transferable to Western Canada and that cost-effectiveness of risk reduction measures should play a larger role in risk management decision-making. Ultimately, we hope that this synthesis and proposed tools will lead to more rational and consistent decisions that save lives and resources in landslide-prone regions of the world.

Section 1: Background information

Methods of expressing societal life loss risk

Life loss risk can be defined simply as the probability of fatality. Life loss risk due to landslides is commonly assessed both for individuals and groups of people. Individual risk is the probability that a specific individual will be killed by a landslide. This risk is often assessed for the individual most at risk within a landslide hazard zone or building and is expressed as the probability of death to an individual (PDI). Individual risk is an important parameter for landslide risk management but is not the focus of this article.

Risk to a group of people is often called “societal risk.” To some, societal risk is a broad concept that considers a variety of widespread and large-scale consequences and provokes a sociopolitical response (HSE 2001, Ball and Floyd 1998, Fell et al. 2005). However, in practice, at least for landslide risk management decisions in Western Canada, societal risk refers more narrowly to the relationship between the probability of, and number of, people killed. Societal risk can be expressed using an FN diagram (Fig. 2), an fN diagram (Fig. 3), or as probable life loss (PLL) (Fig. 3). Figure 4 is a hypothetical example that demonstrates how these methods are related. These methods are used in practice for a diverse range of hazards, including, for example, dam breaches (ANCOLD 2003; USBR 2011; FERC 2016), chemical plant explosions and radiation from nuclear power plants (Ball and Floyd 1998), flooding (Jonkman et al. 2011), and landslides (ERM 1998; AGS 2007; Porter and Morgenstern 2013).

Fig. 3

An fN diagram, showing annual probable life loss (APLL) contours

Fig. 4

A hypothetical example of societal landslide risk at a debris flow fan where risk is “unacceptable” based on the Hong Kong risk evaluation tool shown in Fig. 2

1. fN

The x-axis of an fN diagram (called “little f-N”) represents the number of fatalities (N) in a particular landslide risk scenario. The y-axis represents the probability (f) of the landslide scenario. Each risk scenario is plotted as a point on the fN diagram to display its relative importance. Plotting risk scenarios on an fN diagram can alone inform many risk management decisions, including:

• Which scenario contributes most to total risk?

• Which scenario should be the focus of risk management actions?

• What is the relative priority of various risk scenarios or hazard sites?

• How much risk reduction is achieved by the proposed mitigation measures?

As such, the fN diagram is an important, practical tool for prioritizing landslide scenarios and hazard sites and designing mitigation measures.

2. PLL

The product of probability (f) and number of fatalities (N) yield probable life loss (PLL). PLL describes the expected number of deaths over a period of time. The PLL from a single risk scenario (i.e., a single fN pair) can be compared against other scenarios, or PLL from various independent risk scenarios can be summed to yield the total PLL for an assessed landslide or study area. If probability is presented as an annual probability, then it is the annual probable life loss (APLL). All PLL values are an abstraction, and for landslides in Western Canada, APLL is almost always a fractional value, because in most risk situations there are many years without a landslide followed by an event that would result in multiple fatalities. The PLL value also hides information, as it can include infinite combinations of possible scenarios and does not differentiate between high-probability/low-consequence and low-probability/high-consequence scenarios. Nevertheless, as long as the PLL is viewed within the context of the risk scenarios it represents, PLL is a valuable tool for cost-benefit analysis and comparison of mitigation options (e.g., risk “before” and “after” mitigation), because it can present all the combined risk scenarios as a single value. When this single value is a small fraction, it can be helpful to present PLL over a longer time period, such as the design life of a structure (e.g., a 100-year period), because this timescale is more likely to include a landslide impact and it often translates PLL values into a whole number that is more easily understood.

3. FN

An FN diagram (called “big F-N”) displays the probability of N or more fatalities. Risk scenarios are combined to form a single curve that is a complementary cumulative distribution function (Baecher et al. 2015). FN diagrams have become the preferred tool in Western Canada for comparing estimated societal landslide risk to tolerable and acceptable risk thresholds (e.g., Fig. 2). FN diagrams are used because, similar to the PLL, all of the related risk scenarios are combined to form a single curve, but in contrast to the PLL, curves plotted on an FN diagrams illustrate the differences between high-probability/low-consequence and low-probability/high-consequence risk.

Risk evaluation criteria

Risk evaluation criteria used by analysts and regulators across fields tend to fall into three categories (adapted from HSE 2001):

Equity-based criteria focus on the level of risk and are based on the idea that all individuals have unconditional rights to a certain level of safety. Risks that exceed that level of safety are intolerable and need to be reduced to a tolerable level, regardless of cost.

Utility-based criteria focus on the possible risk reduction measures. Risk reduction measures are needed if the benefits of the measures are greater than the costs of introducing them. Costs and benefits (e.g., statistical lives saved, natural capital protected) are described in monetary terms and compared. Balance can be intentionally skewed by specifying gross disproportion between costs and benefits.

Technology-based criteria focus on the state of practice for risk management. A tolerable level of risk prevention is attained when “state-of-the-art” or “state-of-practice” risk reduction measures are implemented. This is often presented in the form of design standards, such as a minimum factor of safety or a specified design event return period.

In geotechnics, many hazards (e.g., flooding, slope failure, settlement) are managed using technology-based criteria, by demonstrating that the specified factor of safety is achieved against prescribed loading conditions (e.g., CDA 2013). Other industries, such as healthcare and environmental protection, use cost-benefit analysis more widely, which is the backbone of utility-based criteria. For example, in the USA, federal agencies have been required to conduct a cost-benefit analysis to inform policymakers and the public of impacts of all major regulations since the 1980s (USEPA 2014). Bründl et al. (2009) explain that natural hazards in Switzerland are evaluated using an equity-based risk tolerance threshold for individual risks, but societal risks are evaluated using utility-based cost-benefit criteria and the standardized EconoMe method (FOEN 2020).

FN criteria referenced for landslides (e.g., Fig. 2) incorporate all three of the risk evaluation criteria categories (ERM 1998). The unacceptable and broadly acceptable zones are both equity-based. The unacceptable zone identifies risks that are intolerable and need to be reduced regardless of cost, and the broadly acceptable zone identifies risks that can be accepted without further risk reduction (Fig. 5). The tolerable risk line separates the unacceptable zone and the ALARP zone. The example from Hong Kong (Fig. 2) is anchored on 10 or more fatalities (N) at an annual probability of 1 in 10,000 (1E-4), with a slope of −1 in log-log space (Ball and Floyd 1998). On an fN diagram, this threshold would correspond to an APLL of 1E-3. The ALARP zone identifies risks that are tolerable if actions are taken to reduce risk to as low as reasonably practicable (ALARP). The ALARP principle incorporates elements of technology-based and utility-based criteria. Risk is ALARP only if best practices are used to manage risk and cost-benefit assessment demonstrates that the cost of further risk reduction is grossly disproportional to the benefit gained.

Fig. 5

Hong Kong societal landslide risk evaluation tool, illustrating landslide risk aversion and justified expenditure on mitigation costs. The “Intense Scrutiny Region” is grouped with ALARP for this illustration because it is also a transitional zone between “unacceptable” and “broadly acceptable.”

In general, risk tolerance within a society varies considerably from one hazard type to another (e.g., consider the 1000-to-1 spending gap on terrorism versus gun violence in the USA, Cohen 2012). The equity-based thresholds and requirement for gross disproportionality inherent to the landslide risk evaluation tool shown in Fig. 5 imply an aversion to “landslide” hazards compared with other hazard types that are managed without these strict requirements. As such, the FN criteria shown in Fig. 5 advocate for preferential allocation of risk reduction effort, funding, and resources to landslide hazards.

ERM (1998) provides a cost-benefit analysis methodology that can be used to assess cost-effectiveness of mitigation measures. “Gross disproportionality” is demonstrated by estimating a maximum justifiable cost (MJC) of mitigation:

$$ \mathrm{MJC}=\mathrm{PLL}\times \mathrm{AF}\times \mathrm{VSL} $$

(1)

where MJC is the maximum justifiable cost of mitigation (cost/time period), PLL is the probable life loss (deaths/time period), AF is an aversion factor (unitless), and VSL is the value of a statistical life (cost/death). Mitigation options that cost less than the MJC would be “reasonable” (at least economically) to implement, and options that cost more than the MJC could be rejected based on poor cost-effectiveness.

The time period of MJC is set by the time period for which the PLL is calculated. For example, if an annual probable life loss (APLL) is used, then the MJC is the maximum annual justifiable mitigation expenditure. If the probable life loss over 100 years (APLL × 100) is used, then the MJC is the maximum justifiable 100-year life-cycle cost for the mitigation.

The aversion factor defines the disproportionality between mitigation costs and risk reduction benefits for cases in which life loss risk reduction is the only assessed benefit. An AF of 1 would mean that the costs and benefits (in terms of life loss risk reduction) of mitigation are balanced (i.e., proportional). ERM (1998) recommends an AF of 20 for the Intense Scrutiny Region and an AF greater than 1 where risk plots in the ALARP zone within one order of magnitude of the tolerable risk threshold. They also comment that an AF of 20 is used in ALARP assessments for potentially hazardous installations in Hong Kong. HSE (2020) suggests that an AF (called “proportion factor” by HSE) of 10 should be used for risks in the ALARP zone near the tolerable threshold, reducing to 1 near the limit of acceptability. There is no broad consensus on an appropriate AF, and according to Baecher et al. (2015), the concept of disproportionality has been slow to penetrate regulatory practice outside of the UK and Hong Kong.

The value of a statistical life (VSL) is a measure of the benefit of risk reduction. Despite its name, VSL is not a value placed on an individual human life. It is an amount that individuals in a given society are willing to pay for increased safety (i.e., small reductions in their risk of death) (Andersson and Treich 2011). For example, a mitigation measure that protects 100,000 people and reduces annual individual risk by 1 in 100,000, on average, could be described as saving “one statistical life.” If those individuals, on average, were willing to pay $100 each for the mitigation, then the value of the statistical life would be $10 million ($100 per person × 100,000 people) (USEPA 2014). VSL is a concept that is widely used for cost-benefit analysis, particularly by government agencies in the USA, Canada, and the UK (Viscusi and Aldy 2003). VSL changes with time, and VSL varies widely around the world (from approximately $60,000 to $18 million) based on wealth and income elasticity. However, VSL (in 2017 US dollars) in Hong Kong ($7 million) is similar to the USA ($10 million) and Canada ($8 million) (Viscusi and Masterman 2017).

Cost-effectiveness of a specific mitigation measure in achieving an increment of life safety risk reduction can be presented in terms of the cost to save a statistical life (Leroi et al. 2005):

$$ \mathsf{CSSL}=\mathsf{MC}/\mathsf{SLS} $$

(2)

where CSSL is the cost to save a statistical life, MC is the mitigation cost (cost/time period), and SLS is the statistical lives saved by the mitigation measure (i.e., difference in PLL before and after mitigation).

As with Eq. 1, the time period of MC and SLS need to be consistent. For example, if MC is presented in terms of a 100-year life-cycle cost, then SLS needs to be lives saved in 100 years. CSSL is also not a value placed on human life. It describes the efficiency of a measure in reducing risk. The mitigation measures with the lowest CSSL are the most efficient options for reducing risk, and a mitigation system would typically be comprised of some combination of the most efficient measures, which can include non-structural measures like community education and monitoring and warning systems.

Origins and current uses

The history and development of societal risk evaluation tools is well documented in the literature. This section highlights advancements that led to the first societal risk evaluation criteria for landslides, developed in Hong Kong (Fig. 2), and subsequent adoption in other locations and industries.

Whitman (1984) summarized societal risks for natural hazards and a variety of engineering projects on fN diagrams and suggested that these risks have been accepted by the public and therefore may give some indication of allowable risk (following the logic of Starr 1969). Whitman proposed an “accepted” risk line based on the risk from contemporary engineering projects, suggesting that a relatively high-risk level (APLL of 1E-2) was tolerated by the public. Later, authors (Slovic 1987; Fell 1994; Finlay and Fell 1997) noted that this may not be a valid method of assessing public risk acceptance because the public was unaware of the risks and options for reducing risk.

Fell (1994) proposed the use of quantitative risk assessment methods and risk evaluation tools for landslide hazards and included a discussion of possible risk evaluation criteria. He hypothesized that the public may accept relatively high risk for natural landslides, similar to Whitman’s accepted risk line, but that the public may require lower risks for human-made slope stability structures and where natural landslides have been remediated.

Finlay and Fell (1997) presented a study of landslide risk perception by residents of Australia and Hong Kong that was intended to inform landslide risk evaluation criteria and included a summary of risk evaluation criteria in use at the time on an fN diagram. The tolerable risk thresholds varied significantly, from a tolerated APLL of 1E-1 by Union Railway (UK) to a risk reduction objective APLL of 1E-5 for potentially hazardous installations in Hong Kong but were centered around a tolerable risk line with APLL of 1E-3. The survey of residents concluded, in part, that APLL of 1E-5 was perceived to be “low” risk and generally accepted, while APLL of 1E-3 was perceived to be a tolerable risk. The FN criteria developed around this time in Hong Kong (Fig. 2) were consistent with these conclusions.

Ball and Floyd (1998) provide a comprehensive summary of societal risk evaluation history and tools prepared for the Health and Safety Executive (HSE) in the UK. It focuses on societal risks associated with hazardous installations, nuclear power stations, and transport of dangerous goods rather than natural hazards but provides important information that is applicable to landslide risk management. Ball and Floyd (1998) did not recommend a specific risk tolerance threshold, but rather summarized those that are referenced by a variety of agencies. Notably, they highlighted that APLL of 1E-1 is tolerated by HSE for communities living near dangerous goods transport routes, such as sea ports, and that APLL of 1E-2 is tolerated in the UK for nuclear installments, given the risks are reduced to as low as reasonably practicable. They noted that other agencies, for example, in Hong Kong and the Netherlands, have selected more risk-averse limits that would appear to promote greater safety but cautioned that the Dutch thresholds have encountered implementation difficulties and have been relaxed in practice. Ball and Floyd’s (1998) strongest recommendation was that FN criteria lines should be seen as guidelines rather than strict thresholds because of the inherent complexity and uncertainty of both quantitative risk assessment and public risk perception.

The Hong Kong Geotechnical Engineering Office (ERM 1998) proposed the first societal risk evaluation criteria used specifically for landslides. The criteria are based on those adopted previously in Hong Kong for “potential hazardous installations,” such as factories that handle dangerous chemicals, and were developed following decades of public outrage at fatalities caused by landslides from engineered slopes (Malone 2005). Two options were proposed, including that shown in Fig. 2, and an option they preferred over Fig. 2, which replaced the “broadly acceptable” zone with ALARP. Risk tolerance thresholds were placed on an FN diagram, with tolerable risk coincident with an APLL of 1E-3. ERM (1998) highlighted that risk plotted on an FN diagram is affected by scale and defined the length of natural terrain (500 m) that was intended to be the basis for risk comparison to societal risk criteria. This natural terrain length was selected so that the population at risk from the landslide (typically within densely developed residential high-rise buildings) was similar in scale to the population at risk from an industrial plant. ERM (1998) chose to adopt risk tolerance thresholds that had been developed for “human-made” hazards because they suspected that the public in Hong Kong would perceive a strong element of government involvement in the decision to allow a development in a landslide hazard zone. ERM (1998) recommended that the societal risk criteria should not be mandatory and should be used as a guideline only. Furthermore, ERM (1998) proposed the criteria as a starting point for extensive consultation, noting that further work was needed to assess public perception, practicality, and cost-benefit implications.

Hungr et al. (2016) provided a summary of landslide risk acceptability practices in various countries, suggesting that the Hong Kong FN criteria for societal landslide risk have been adopted and widely referenced only in Hong Kong, Australia (AGS 2007), and Western Canada (Porter and Morgenstern 2013). Other countries (e.g., Europe, Japan) manage landslide risk using landslide hazard probability thresholds or individual risk tolerance thresholds (e.g., Corominas and Mavrouli 2010).

FN and fN risk tolerance criteria have also been widely adopted by the hydroelectric and flood-control dam industries, particularly in Australia, the USA, and Canada. Baecher et al. (2015) provided a comprehensive summary of societal risk criteria for loss of life focusing primarily on dams and levees, including a detailed description of the mathematical concepts and history. In Australia, societal risk related to dam breach is plotted on an FN diagram and compared to a tolerable risk line that is equivalent to the Hong Kong FN criteria for landslides (APLL of 1E-3) (ANCOLD 2003). In the USA, societal risk related to dam breach is plotted on an fN diagram (USBR 2011; FERC 2016) and compared to a “reference line” with APLL of 1E-3. Terms like “risk tolerance criteria” and “tolerable risk” are intentionally omitted in recognition of the inherent uncertainties in risk assessments and risk perceptions and the many other factors (not related to life loss risk) that influence a decision. The reference line separates zones of increasing and decreasing justification to reduce or better understand the risk.

Hong Kong FN criteria summary

This review highlights the following characteristics of the Hong Kong risk evaluation tool for landslides (Fig. 2):

• 1. Landslide risk aversion: The position of the tolerable risk line (APLL = 1E-3) and the aversion factors proposed by ERM (1998) are both based on the assumption that Hong Kong’s society is more averse to landslide hazards than other natural hazard types, implying that the society ought to invest more resources in mitigating landslide hazards than other hazard types (Fig. 5). Also, Ball and Floyd (1998) showed that higher levels of risk are tolerated by the public in the UK (using equity-based criteria) for transport of hazardous goods, chemical plant installations, and nuclear plants. The aversion factor and “gross disproportionality” test of the ALARP principle require that cost-benefit balance is skewed heavily towards high expenditure for modest landslide risk reduction, which could not be sustained for all hazards that are faced. It appears to be designed for a society with large financial resources that is affected disproportionately by landslide hazards.

• 2. Human-made hazard: The aversion to landslide hazards is, in part, due to the interpretation that landslide hazards are perceived to be a “human-made” technological hazard, similar to a chemical plant or an engineered structure, such as a hydroelectric dam.

• 3. Scale dependence: fN and FN diagrams can be used to plot any number of scenarios, affecting any number of people. However, the risk tolerance thresholds superimposed on these diagrams are affected by scale, and the criteria may not be directly applicable at other scales. The x-axis (number of fatalities) is affected by the total number of people exposed to the hazard. The y-axis (probability of fatalities) is affected by the range of event probability (i.e., return period) considered. The Hong Kong landslide risk evaluation tool was developed for an area affected by a single slope segment and considering up to a notional 1000-year return period event for routine assessments (Ho and Roberts 2016).

• 4. Interim guideline: The Hong Kong risk evaluation tool was a first proposal that requires review and possible revision and was intended to be a reference for decision-making, rather than a strict threshold for risk acceptability. ERM (1998) recommended further work to assess applicability of the evaluation tool and noted that practical problems with the evaluation tool could only be identified following extensive use.

Section 2: Case studies in Western Canada

This section describes our experience with societal landslide risk evaluation for residential development in Western Canada, including what has been learned about the utility and challenges of applying FN criteria.

Residential development in Western Canada has historically (particularly prior to the 1970s) occurred without adequate recognition of landslide hazards, and few existing developments in hazard zones are effectively protected. The risk management process is usually triggered by a landslide event, often in areas that were not fully aware of the threat to their community. A provincial code, standard, or specification for landslide risk management does not exist, and there is no national nor provincially adopted level of landslide safety in British Columbia or Alberta. Hazard and risk management for residential development is delegated primarily to the local government, and each local government manages hazards in its own way. Landslide hazards are assessed at proposed developments (e.g., following guidance in Cave 1992; APEGBC 2010), but there are no legislated triggers for assessment and evaluation at existing developments.

A few municipal governments, such as District of North Vancouver (DNV) and Town of Canmore (TOC), have responded to landslide and debris flood events by developing local regulations for assessing hazards and managing risk using a quantitative risk management (QRM) process (Fig. 1) (VanDine 2012). The QRM process is used to support land use decisions and applications to solicit funding for mitigation measures. QRM was introduced to these communities by engineering and geoscience consultants, and the risk evaluation tools that have been referenced are modelled after those developed in Hong Kong for landslide hazards (Porter and Morgenstern 2013). Other local governments are now increasingly referencing FN criteria and regulations adopted by DNV and TOC, and during the last 10 years, FN criteria that originated in Hong Kong (Fig. 2) have had a strong influence on risk management decisions at more than 50 landslide and debris flood hazard sites across British Columbia and Alberta.

A sample of these landslide and debris flood hazard sites for which we have baseline risk estimates, cost estimates for proposed mitigation measures, and risk estimates following mitigation (i.e., residual risk) is summarized in Table 1 and Fig. 6. Although each hazard site is unique and risk management decisions are subject to site-specific constraints, objectives, and uncertainties, the following patterns have emerged:

• 1. All evaluated sites require mitigation: Only development sites that are within a landslide hazard zone tend to be selected for a detailed quantitative risk assessment, and in most cases the landslide is unmitigated and unmonitored. The typical practice in Western Canada has been to assess up to at least the 2500-year return period event, and occasionally up to the 10,000-year return period event, which represents the upper bound magnitude in the post-glacial record (Jakob et al. 2018). The intensity of such large events means that societal risk invariably plots in the unacceptable or ALARP zones of the FN criteria, even if only one building is impacted.

• 2. Required mitigation is expensive: The risk reduction designs are sized to meet the risk tolerance objectives directed by the FN criteria, either without regard for cost or seeking designs that are disproportionately costly (Fig. 5). Although there is no defined policy, decision-makers have tended to specify that risk be reduced by one zone on the FN diagram. That is, unacceptable risks need to be reduced to ALARP, and risks in the ALARP zone need to be reduced to broadly acceptable or so that N is less than 1. Where risk avoidance is impractical, risk reduction measures have typically included protective structures (e.g., diversion berms and channels, barriers, basins), and to meet the owners’ risk reduction objective, these structures typically need to be designed to manage the largest event considered in the risk assessment (which tend to be 2500 to 10,000-year return period events). Estimated costs for risk reduction measures tend to be on the order of several million to tens of millions of dollars (CAD) (Fig. 6).

• 3. Mitigation funding is difficult to obtain: Local governments responsible for managing the hazards typically have a strong political mandate to protect their citizens, but they rarely have the financial resources to fund structural mitigation measures, particularly the expensive measures dictated by the risk evaluation tool. Local governments compete for funding from a variety of provincial and federal grants (which can take years without a guarantee of success), and the grants often have a maximum value that is insufficient and unrelated to the cost of reducing risk to a tolerable level. Winning a grant is a function of many factors (e.g., number of applicants, timing of submittal, availability of grant money) that may be unrelated to urgency of need for a particular community, and ultimately there is not enough funding to meet the requests of all applicants. Several local governments are partnering with land developers, allowing new development in landslide hazard zones if the developer pays capital expenses for mitigation that protects both existing and new development. Others have decided to not fund future landslide risk assessments because the expensive assessments have not resulted in risk reduction funding.

• 4. Wealthy communities are more likely to acquire funding: Larger, affluent local governments tend to be more successful at acquiring mitigation funding than poorer communities (Table 1). These governments have larger municipal budgets, which allows for direct funding of mitigation measures and cost matching agreements with provincial and federal funding sources and tend to have more resources, knowledge, and experience that lead to successful mitigation grant proposals.

• 5. Economics, rather than life loss risk, drives funding decisions: Projects with high economic risk (e.g., building damage, loss of business) tend to be funded before those with low economic risk and high life loss risk. This may be because funding proposals can clearly demonstrate the risk reduction benefit (typically showing a positive return on investment) when risk reduction is described in economic terms. Alternatively, or in addition, it may be that high economic risk tends to occur in wealthy, rather than poor, communities.

• 6. Life loss risk reduction is not yet achieved at many sites: Many of the hazard sites assessed in the past decade that are driven by life loss risk (with relatively low economic risk) have not received mitigation funding and there have been no physical actions to reduce risk. Some communities have had opportunities to receive lesser amounts (less than $1 M) of funding but have not known how to spend it because the mitigation strategies focused on expensive measures designed to meet FN risk tolerance objectives, rather than low-cost risk reduction measures, such as monitoring and warning systems or less robust protection structures.

• 7. Other important factors have not influenced funding decisions: There are several factors that appear important but have had little influence on funding decisions, including:

  • a. High life loss risk: Mitigation funding has not been granted to the highest risk cases. For example, only two out of six “high” risk (> 1 statistical life saved in 100 years) cases shown in Fig. 6 have received funding.

  • b. Multiple fatality events: The theorized aversion to multiple fatality events discussed widely in risk management literature has not led to mitigation funding. The size of each point in Fig. 6 is scaled to the estimated number of fatalities in an extreme event. The larger points, which indicate more potential fatalities in a single event, tend to be “not funded.”

  • c. Cost-effectiveness: The cost-effectiveness of mitigation measures in terms of CSSL or as a function of VSL has not typically been assessed or reported, partly because of the focus on reducing risk without regard for cost. Cost-effectiveness has apparently also not been a factor considered by funding authorities. For example, four out of seven cases with “good” cost-effectiveness (i.e., CSSL < VSL) are “not funded” (Fig. 6).

Table 1 Landslide and debris flood risk management case histories in British Columbia (BC) and Alberta (AB) that referenced FN criteria, ordered by date of study

Fig. 6

Landslide risk management case histories in Western Canada. Labels correspond to Table 1. ($) symbol indicates sites where economic risk is a key driver. Size of each point is scaled to “extreme event fatalities (N)” shown in Table 1. The cost to save a statistical life (CSSL) is greater than the value of a statistical life (VSL) in the dotted red zone

In summary, at many landslide hazard sites in Western Canada, the adoption of strict (i.e., risk averse) societal risk tolerance criteria (Fig. 2) has not led to greater risk reduction. In the worst cases, the criteria have led to mitigation designs that are unaffordable or inefficient. But more often, FN diagrams, on their own, have simply not been an effective tool for informing funding authorities that the particular landslide risk management solution is a justified allocation of a government’s limited resources.

Section 3: Proposed societal risk evaluation tools

Societal life loss risk assessment and FN criteria are valuable decision-making tools. However, the challenges described in Western Canada suggest that adaptation of the landslide risk evaluation tools may be needed in each sociopolitical context, which supports conclusions drawn previously by Macciotta and Lefsrud (2018). Stated plainly, risk evaluation tools that are highly effective in Hong Kong may be less effective (or potentially inappropriate) elsewhere because of differing attitudes towards risk, perceptions of landslide hazards, financial constraints, funding mechanisms, and risk management frameworks (e.g., maximum event return period included).

Every society has limited resources to manage a wide range of hazards, including (naming only a few) traffic accidents, violence, environmental pollutants, pandemics, earthquakes, food scarcity, and fires. A goal of this article is to take one small step at improving resource allocation across all hazard types, by advocating for fair and efficient allocation of resources to landslide hazards. With that goal in mind, this section describes opportunities to improve societal life loss risk evaluation tools. The tools are calibrated towards the experience in Western Canada but may also be a useful starting point for other societies.

Principles

The proposed tools are based on the following broad principles:

• 1. Transparent risk evaluation tools are needed: Decision-makers need a framework for communicating “why” each decision is made. This is the purpose of risk evaluation criteria. Technology-based criteria, such as a minimum factor of safety, are typically not appropriate for landslide hazards from natural slopes. Therefore, an equity-based (i.e., risk tolerance threshold) or utility-based (i.e., cost-benefit threshold) tool is needed.

• 2. Aversion to landslides over other hazards is not appropriate: Aversion to landslide hazards compared with other hazard types (implying that mitigation resources should be directed away from hazards like pandemics, earthquakes, or traffic accidents and towards landslides) is not justified in Western Canada, nor in most other places. Case histories presented in “Section 2: Case studies in Western Canada” do not suggest preferential allocation of resources to landslide mitigation in Canada. Additionally, research by Viscusi (2009) suggests the opposite, concluding that Americans are willing to pay twice as much to prevent a traffic accident death compared with a natural disaster death. The aversion to landslide hazards that is inherent to Hong Kong’s landslide risk evaluation tool (Fig. 2) can have hidden costs; it can divert funding away from other, greater hazards and/or it can lead to unaffordable landslide mitigation designs. There is an opportunity cost for overspending on landslides. For example, in the healthcare industry, Broughel and Viscusi (2017) estimate that lifesaving regulations that cost more than $100 million USD (CSSL that is 10 times the VSL) are counterproductive and cause more expected fatalities than they prevent.

• 3. Landslide hazards from natural slopes are not generally perceived as “human-made”: In their presentation of proposed FN criteria for Hong Kong, ERM (1998) contended that risk evaluation criteria developed for human-made, technological hazards (e.g., chemical plants) were appropriate for landslides in Hong Kong (Fig. 2) because residents viewed the government as responsible for landslide disasters given that government approved the development locations. It is also notable that many of the landslide disasters in Hong Kong prior to that time involved engineered slopes and retaining walls (Malone 2005). In contrast, the compilation of Wachinger et al. (2010) suggests that Hong Kong is an outlier, noting that most societies tend to tolerate higher levels of natural hazard risk compared with technological or human-made hazards. Our experience also suggests that perceptions of natural hazard risk at existing development in Western Canada may be different from Hong Kong. Specifically, as shown in Fig. 6, many hazard sites that plot in the “unacceptable” zone of the FN diagram remain “not funded” by authorities, and residents have generally remained in their homes. We observe in Western Canada that higher risks tend to be tolerated for newly recognized landslide hazards affecting existing development, than for new developments proposed in previously recognized landslide hazard zones.

• 4. Emphasis on cost-effectiveness is needed: Cost-benefit ratio and cost-effectiveness of potential mitigation measures should be primary landslide risk evaluation considerations. Utility-based risk evaluation thresholds (e.g., required cost-benefit ratio) can be presented as transparently and simply as risk tolerance thresholds and are standard practice for natural hazard risk management in Switzerland (Bründl et al. 2009; FOEN 2020). The great advantage of utility-based thresholds is that they are implicitly benchmarked against other hazard types and locations, which contributes to justifiable and efficient allocation of resources. They distinguish mitigation proposals that are clearly “good investments” from those for which decision-makers should pause and consider opportunity cost, asking: “Would this money be better spent elsewhere?” Finally, and maybe most importantly for the pragmatists, they bring mitigation proposals fully into economic terms, which appears to facilitate funding decisions, at least in Western Canada.

• 5. A resource-constrained objective is needed: In our experience in Western Canada, the stated objective of many landslide risk management projects has been to “reduce risk to tolerable levels.” Although noble, Fig. 6 suggests it is commonly not achievable. That objective is challenged by practical constraints, scarce resources, competing objectives, and competing hazard types. A more pragmatic objective is often needed, such as to “reduce risk to the extent possible with available resources.”

• 6. Improved risk communication is needed: Clear communication is needed to allow community leaders and the public to make informed decision about risks they face and actions to take. Clear communication involves speaking and listening. Landslide risks need to be presented in everyday language and in context with other hazards. People generally want to understand if the landslide risk is important relative to other hazards, the potential scale of impacts to their communities, and how they can participate in improving the resiliency of their community.

Proposed tools

The tools shown in Fig. 7 and Fig. 8 are proposed for societal landslide risk evaluation in Western Canada and may be applicable in other similar societies. The authors welcome discussion and recognize, like ERM (1998), that ongoing review is still needed to assess their effectiveness. The tools are based on the following suggestions to landslide risk analysts and decision-makers who face landslide hazards:

• 1. Avoid risk assessment conservatism: Engineers are trained to incorporate conservatism into their design assumptions. However, risk is overestimated when this conservative attitude is applied, perhaps unknowingly and to a number of different inputs, in a risk analysis. Inflated risk estimates are inappropriate for risk evaluation. Analysts should assess and present uncertainties transparently while using best estimates for risk evaluation.

• 2. Evaluate individual risks: Societal risk evaluation tools, particularly those focused on risk tolerance thresholds (e.g., Fig. 2), were intended for hazard scenarios with large-scale consequences that provoke a sociopolitical response. In terms of life loss, “large scale” was commonly regarded as 10 or more fatalities (Ball and Floyd 1998; PSC 2020). Societal risk tolerance thresholds do not need to be referenced in cases where individual risk tolerance thresholds suffice for decision-making. Societal risk evaluation tools are more difficult to explain to the public than individual risk tolerance thresholds and can contribute more to confusion than a transparent decision.

• 3. Use guidelines rather than thresholds: Lines shown on fN and FN diagrams should be viewed as references to inform decisions and provide justification for action (or inaction) rather than strict criteria. This recommendation is often repeated by the creators of societal risk evaluation tools (Ball and Floyd 1998) but is sometimes neglected by practitioners. The U.S. Bureau of Reclamation (2011) provide a description of the distinction between a criteria and guideline. The reference line included in Fig. 7 (APLL of 1E-3) is consistent with the Hong Kong risk tolerance threshold and the reference lines used by dam agencies in the USA, Canada, and Australia. An individual risk tolerance threshold is generally applicable in cases where a strict criterion is needed.

• 4. Benchmark to other landslide sites and hazard types: An understanding of the relative level of risk at various landslide hazard sites and other hazard types within a region or administrative body is needed to set guidelines on FN diagrams and, more broadly, for effective resource allocation. We are involved in ongoing work to better understand levels of societal landslide risk in the regions of Western Canada for this purpose and hope that it can better inform the position of FN guidelines at some point in the future.

• 5. Cost-effectiveness considerations and standard of practice always apply: The utility-based and technology-based criteria utilized for hazard sites that plot in the ALARP zone (Fig. 2) should be applied to a wider range of projects. Cost-benefit and cost-effectiveness concepts should inform all risk management decisions, and all newly implemented mitigation designs should be consistent with the standard of practice. For example, low-cost monitoring programs (e.g., periodic lidar change detection) may be a cost-effective measure to inform management of low-probability/high-consequence landslide risk scenarios (e.g., a rock avalanche from a natural slope).

• 6. Recognize scale effects: The reference line included in Fig. 7 was developed for an area affected by a single slope segment that threatens hundreds or thousands of people and considering up to a notional 1000-year return period event (Ho and Roberts 2016). It is less relevant at other scales, such as for hazard scenarios that threaten millions of people (e.g., regional flooding, hurricane) or have a return period greater than 1000 years.

• 7. Gross disproportionality is inappropriate: Preferentially allocating resources to landslide hazards compared with other hazard types is not generally appropriate. Resources should be allocated to hazard sites in proportion to the level of risk. VSL and CSSL are important concepts for assessing proportionality.

• 8. Broadly acceptable and intense scrutiny zones are not needed: Few hazard sites plot in these zones on an FN diagram, and they generate confusion when FN diagrams are communicated to the public. Item 5 (above) describes considerations for low-probability/high-consequence scenarios that are the focus of the intense scrutiny zone.

• 9. A higher standard is needed for new development: A higher standard is likely appropriate for new developments that are placed in recognized hazard zones. New development without mitigation increases risk, and there is an obligation to manage this risk proactively.

• 10. Use societal risk evaluation tools for mitigation design and budgeting: Societal life loss risk expressed in terms of PLL or fN can be a powerful tool for assessment of mitigation options, selection of mitigation design event, allocation of budget across multiple hazard sites, and evaluating tolerability of increased population density in a hazard zone. Figure 8 is a tool for these purposes. It illustrates the maximum justifiable cost (MJC) of mitigation (Eq. 1) for an aversion factor of 1 (no special aversion to landslide hazards), mitigation cost in terms of the net present value of the 100-year life-cycle cost, and value of a statistical life (VSL) of $10 million. VSL varies with time and place (Viscusi and Masterman 2017), so the tool will need to be updated if this VSL estimate is an unrealistic estimate for the intended project location. The tool could be used for preliminary budgeting by plotting total risk in terms of fN for each hazard site to identify a reasonable mitigation investment per site. This simple understanding of reasonable expenditure can help to focus the conceptual stage of mitigation design and mitigation options comparison. Also, multiple scenarios at a single hazard site can be plotted to understand the relative contribution of each scenario to total risk. This approach can guide selection of components in a mitigation system and selection of the design event used to size mitigation elements. Two examples of the use of this tool are shown in Fig. 9.

Fig. 7

Proposed FN diagram for societal landslide risk evaluation in Western Canada, intended for ongoing review and discussion. Note the shift in extent of both the x- and y-axis compared with Fig. 2. The reference line is equivalent to the tolerable risk threshold shown in Fig. 2.

Fig. 8

Proposed fN tool showing the maximum justifiable cost (MJC) for mitigation that results in cost-effectiveness (CSSL) similar to the value of a statistical life (VSL). This tool is specific to mitigation cost in terms of net present value of 100-year life-cycle cost, VSL of $10 million, and AF or 1. It requires modification for other conditions. PLL100 is probable life loss over 100 years

Fig. 9

Proposed fN tool showing example scenarios and hazard sites. Numbered sites correspond to the examples described within. Lettered sites correspond to case histories in Table 1 and Fig. 6

Conclusions

Every society is resource-limited, with more needs than resources and more hazards than solutions. Quantitative risk management is a valuable framework in this environment because it allows the multitude of hazard types and locations to be compared and prioritized, leading (in principle) to well-informed allocation of resources. Risk evaluation is a critical step in this framework; it is where leaders decide that the people are “safe enough” or how much protection is needed. The societal landslide risk evaluation criteria developed in Hong Kong were a great advancement for this purpose and remain unrivaled.

After adoption and more than a decade of application in Western Canada, it is becoming clear that the societal landslide risk evaluation tool designed for Hong Kong may not be directly transferable to places with lower population density, larger landslides, less financial resources, and different perceptions of risk and government responsibility. In retrospect, it is unsurprising that the Hong Kong FN criteria are not universally applicable. After all, each society on earth has unique values, perceptions, and beliefs, and a risk evaluation tool is nothing but a gross simplification of these attitudes.

This article proposes modified societal landslide risk evaluation tools designed to overcome the challenges of applying the Hong Kong FN criteria to Western Canada. Specifically, the proposed tools remove the aversion to landslide hazards compared with other hazard types that is inherent to the Hong Kong FN criteria and place more emphasis on cost-effectiveness of landslide mitigation options. Although the proposed tools were developed for the sociopolitical context of Western Canada, we hope the discussions included in this article can motivate others to modify these tools for other societies and hazard types.