The use of Risk and Vulnerability Analysis in Climate Change Adaptation

Jens Laugesen

Det Norske Veritas, Høvik,Norway

Bodil Aamnes Mostue

SINTEF Technology and Society, Trondheim, Norway

Ingrid Bouwer Utne

Department of Marine Technology,Norwegian University of Science and Technology (NTNU), Trondheim, Norway

Jørn Vatn

Department of Production and Quality Engineering, NTNU, Trondheim, Norway

ABSTRACT

A wide range of risk and vulnerability analysis (RVA) methods exist in the literature and are also used in relation to climate changes. A traditional approach to RVA might be sufficient to assess risks and vulnerabilities on a superior level, but for climate change adaption more detailed analyses are necessary to improve the results. For flooding this involves detailed mathematical and probabilistic modeling of hydrology, surface absorption capacities, snow melting, and failure or breakdown of critical components or system. The challenge is to link the identified threats and vulnerabilities to the risk picture taking physical models and climate projections explicitly into account. The objective of the paper is therefore to extend current RVA methods to combine physical modelswith results from climate projections. In the paper, a typical approach to RVA is applied to a flooding event and challenges associated with the method and the results of the analysis are discussed as basis for evaluating the need for extensions of the traditional RVA. We propose a seven steps method that integrate the dose-response analysis related to flooding into RVA.

1INTRODUCTION

Climate change is on the agenda worldwide. Municipalities have important responsibilities in local climate change adaptation, for example, with respect to assessment of risks related to land use planning and construction in urban areas. Today, risk and vulnerability analysis (RVA) is often used by municipalities and infrastructure owners in Norway as a basis for their emergency preparedness planning and to ensure that public safety are sufficiently integrated in land use planning (DSB 1994).

RVAis a method used to identify and document risks and vulnerabilities in a specific system, and resembles a preliminary hazard analysis (PHA) (Ericsson 2005). The main objective of the analysis is to improve the system’s robustness towards hazards and threats. By combining knowledge from experts, managers and/or users in a structured analysis and decision process, undesired events are identified and ranked, depending on their risks. Then the risk picture for the system can be described and used as a basis for determining risk and vulnerability reducing measures (DSB 1994).

RVA was developed in the early 1990’s as a response to the need for risk assessments based on laws and regulations. At that time, it was recognized that small and medium enterprises and the public sector did not have sufficient resources to carry out extensive analyses, such as probabilistic safety analysis (PSA) and quantitative risk analysis (QRA). Thereafter, RVA has been applied for many years successfully. However, the recent need for climate change adaption to mitigate societal risks, such as flooding and landslides has made it necessary to reconsider RVA as a feasible risk assessment tool for municipalities in their land use planning.

Currently, there are many tools and frameworks suggested fordifferentchallenges related to climate change adaptation, as suggested by UNFCCC (2008),Jones (2001) and Lindley et al. (2006).

For climate change adaption a traditional approach to RVA might be sufficient to assess risks and vulnerabilities on a superior level. However, more detailed analyses involving detailed mathematical and probabilistic modelingare necessary to improve the results. A challenge is to link the identified threats and vulnerabilities to the risk picture taking physical models and climate projections explicitly into account.

The objective of the paper is to extend the current RVA method so that it is possible to combine physical models, e.g., drainage models, with the results from climate projections. The paper starts with describing the basic process of RVA. Then a typical approach to RVA is applied to a flooding event and challenges associated with the method and the results of the analysis are discussed as basis for evaluating the need for extensions of the traditional RVA. Last, a seven step method that integrates the dose-response analysis related to flooding into RVA is proposed.

The work described in this paper has been undertaken as part of two research projects;AdaptCRVA (2007-2009) and FloodProBe (2009-2013), as well as a case study with two municipalities, Stavanger and Sandnes, in Norway. The paper presents results based on experiences from these projects.

2THE RVAMETHOD

RVA can be used as an initial risk study in an early stage of a project, as an initial step of a detailed risk analysis of a system concept or an existing system, or as a complete risk analysis of a rather simple system. All hazards and possible accidental events must be identified. It is important to consider all parts of the system, operational modes, maintenance operations, safety systems, and so on. In general, the RVA is a semi-quantitative methodused to:

  1. Define objective and determine RVA prerequisites, and establish a RVA team consisting of personnel with system knowledge and expertise on risk assessment methodology.
  2. Identify hazards and hazardous events.
  3. Estimate frequencies/probabilities and consequences.
  4. Rank risks, determine risk reducing measures and follow-up actions.

A hazardous event may lead to a wide range of consequences, ranging from negligible to catastrophic. Various types of consequences can be considered, e.g.: life/health, material assets and environment. By using RVA, municipalities can identify relevant hazards or undesired events, such as:

Natural events (e.g. flooding, landslides, extreme weather conditions).

Damage to/loss of critical infrastructure, e.g., energy supply, water supply, information and communication technologies (ICT).

Large traffic accidents

Fires and explosions

Large industrial accidents

Breaking of dikes

Acute pollution

Sabotage and terror

Radioactive fall-out

The risk related to an undesired event is a function of the frequency/probability of the event and the severity of its potential consequences. The assessment of probabilities and consequences of the events may be based on categories, such as the examples shown in Table 1 and 2. Depending on the system and the scope of analysis, other categories may be defined.

Often it is practical to put the different hazardous events into a risk matrix to compare the events, as a basis for prioritizing of further actions, for example by using the As Low As Reasonable Practicable (ALARP) principle (Norsok 2001). Decisions on acceptance criteria and acceptable risks are subject to company or municipality decisions and regulations, and should be established before the analysis is initiated.

Table 1: Probability/frequency categories (example)

Category / Probability / Frequency
1 / Very unlikely / <1/1000 year
2 / Remote / 1/100-1/1000 year
3 / Occasional / 1/10-1/100 year
4 / Probable / 1/1-1/10 year
5 / Frequent / >1/year

Table 2: Consequence categories (example)

Category / Consequence / Life/health / Loss of material assets / Environment
€
1 / Insignificant / Small personal injuries / Less than 25,000 / Small area affected, short restitution time.
2 / Minor / Serious personal injuries / 25,000-250,000 / Large area affected, short restitution time.
3 / Major / 1-2 fatalities / 0.25 – 2.5 mill / Small area affected, long restitution time.
4 / Critical / 3-10 fatalities / 2.5 -25 mill / Large area affected, long restitution time.
5 / Catastrophic / >10 fatalities / >25 mill / Permanent damage.

Even though the risk matrix usually gives an indication of how to prioritize risk reducing measures, it may not be easy to determine which measures should be implemented. As a starting point a cost/benefit approach may be used, but in many cases it may be difficult to determine costs, effectiveness, and benefits. Decisions may depend on stakeholders’ preferences and risk perception. Often, the decisions are results of political processes in which other aspects than safety are taken into considerations.

3Climate projections

Climate projections, also called climate change scenarios, provide a picture of how future climate is expected to be. Expected climate changes in Norway with respect to precipitation, wind and temperature towards 2100 are as follows (Hanssen-Bauer et al. 2009):

Precipitation:

Increase in precipitation in all regions and for all times of the year.

Average increase in annual precipitation between 5 to 30 % towards the end of the century.

The winter precipitation can increase with over 40 % in eastern, southern and western parts of Norway towards the end of the century.

Increase in days with heavy precipitation. The average precipitation for these
days is higher in all regions and for all seasons.

Wind:

Climate models provide little or no change in the mean wind conditions during this century.

The frequency of higher wind forces can increase, but due to systematic weaknesses in climate models, it is too early to conclude about this.

Temperature:

It will be warmer in all regions and for all seasons.

Annual mean temperature will increase by 2.3 to 4.6 °C within 2100.

Highest temperature increase during winter, and least during summer.

Highest temperature increase north, where annual mean temperature is estimated to increase 3.0 to 5.4 °C.

The climate projections are uncertain due to lack of knowledge of future emissions of greenhouse gases and particles, combined with deficiencies and simplifications in the climate models. The uncertainties are larger in local scale than global. Nevertheless, the climate projections give an indication of what kind of changes to expectin the 21 century.

4USE of RVA for climate change adaptation in municipalities

In the following, the typical approach to RVA is applied to a flooding event, as basis for evaluating the need for climate change adaptations of the current method. Thereafter, a method for an extended RVA is proposed.

4.1RVA of flooding

Undesired event: Due to rising sea levels a coastal municipality risks flooding at the housing area “Fjord dream” built on piles in the harbour area of the municipality.

The municipality assumes that the probability for such an event to occur is every 50-100 years, and it is assumed that the probability will increase in the future due to climate change. Using the probability/ frequency categories in Table 1 gives a probability category of 3 (occasional).

The consequencesare evaluated by the municipality and they conclude that (Table 2):

Life/health (category 1– insignificant): It is assessed that the consequence for life/health is insignificant because the area can be evacuated in due time before the houses get flooded.

Material assets (category 4 - critical): The area contains 120 houses which have been constructed for € 0.4 million each. A flood will destroy all the houses and these have to be reconstructed at 120 x 0.4 = € 48 million. It is assumed that it will cost an additional € 10 million to clean up the area before the area can be reconstructed.

Environment (category 2 - minor): It is assumed that the whole housing area in the harbour can be affected, after the area has been cleaned up the area,it will have a short restitution time.

Probability/
Consequence / 1
Veryunlikely / 2
Remote / 3
Occasional / 4
Probable / 5
Frequent
5
Catastrophic
4
Critical / Material assets
3
Major
2
Minor / Environ-ment
1
Insignificant / Life/
Health

Figure 1: Risk assessment matrix for flooding.

The result is that for the material assets,further evaluations should be considered. Based on the ALARP principle,risk reducing measures should be carried out provided that they are not disproportionately large compared to the risk reduction that is achieved.

Two different risk reducing measures are presented to the municipality:

  1. Raising all the piling foundations for the houses by 1 m and thereby reducing the probability for a flooding to less than 1 in every 1000 years (very unlikely). The operation is relatively expensive because every house has to be jacked up to be able to install the extension of the piles. The cost is estimated to be € 60000 pr. house, totaling € 7.2 million.
  2. Build a dike in the sea outside the housing area to protect the houses. The dike will go 1.5 m over the average sea level today and have a marina inside for the boats of the residents. The access to the sea for the boats will be ensured by establishing a sluice gate in the dike. The probability for a flooding with this solution is less than 1 in every 1000 years (very unlikely). The cost for establishing the dike and the sluice is estimated to be € 4.5 million.

As both risk reducing measures reduce the probability to the same value the municipality could decide to go for the dike alternative which has the lowest cost.

When assessing the risk associated with each risk reducing measure we need to consider various climatic scenarios. A first approach would be to use a most likely scenario as a basis, and conduct the analysis based on this scenario. In our situation this means to fix a given sea level increase. The assessment in our example is made based on the most likely increase in sea level. However, if a worst scenario assumption is made, the result will change. In our example it is expected that the return period for flooding will increase for both scenarios, but not necessarily in the same manner. For example if the flooding frequency is considerably higher with a dike outside the housing area compared to raising the piling foundations for the houses in a worst case scenario this gives stronger arguments to choose the piling alternative despite the higher cost. Sensitivity analyses of scenarios and associated cost-benefit analyses are therefore recommended. Sensitivity analyses may be carried out without any consideration of the likelihood for the various scenarios, whereas a full cost-benefit analysis also requires assessing the probabilities for the various scenarios. In the present example we do not present explicit sensitivity analyses nor cost-benefit analyses.

4.2Challenges with the traditional approach to RVA for flooding

The traditional approach to RVA gives in many cases limited decision support. More structuring in collecting and adapting knowledge are often necessary to estimate probabilities and consequences of 1) events when several simultaneous weather conditions occur, e.g. great wave height, storm surge and rising of sea levelin the example above and 2) events reflecting different extent of climate changes, e.g.uncertainty whether the sea level rising is 0.5 meters or 1.5 meter in 100 years. More detailed analyses are necessary to improve the decision support. For flooding thisinvolves detailed mathematical and probabilistic modeling of hydrology, surface absorption capacities, snow melting, and failure or breakdown of critical components or system. Such modeling is similar to well known dose-response models. The “dose” here is the various weather conditions in combination with system failure, and the “response” is how the surroundings absorb precipitation, and how water accumulates and builds up in the waterways.

A dose-response model is in principle a deterministic model implementing a range of physical laws. Uncertainty, and hence risk, comes to play in two different manners: One dimension is the uncertainty regarding parameters in the physical (response) model. For example, it is hard to assess the permeability of the surroundings to the necessary precision. To some extent such uncertainty may be reduced by model calibration. This, however, requires access to weather data, and measurement of the flow situation in the actual waterway. Today, in planning of new residential and industrial areas it is common to establish such models, and to some extent calibrate the models with real observations.

The second type of uncertainty relates to the relevant “dose” scenarios since there is almost an indefinite number of undesired events or scenarios to consider. Hence, there has to be an initial screening when it comes to which types of scenario to include in the model. For example, precipitation is often described by combination of intensity-, duration- and frequency. The combination of intensity and duration is a challenge, as well as to include factors like temperature, wind condition, and snow depth in the analysis. Thus, there is an uncertainty regarding whether the most relevant scenarios have been selected for inclusion in the model.

Next comes the uncertainty regarding the frequencies. To some extent there are historical data available to describe the situation as it is today, but it is necessary to link the projection models for climate change to the dose-response models in order to improve the input to RVA.“New” nature phenomena might occur with significantly changed frequency, strength, in new times of the year or at new sites. Examples of natural disasters that may come to act in unanticipated ways are water saturated landslides, rock falls triggered by the freeze/thaw processes, slush avalanches, precipitation floods in small rivers, urban flooding, and sea level rise (Aaheim et.al. 2009). Groven et al (2008) argue that the use of traditional approaches may not pick up such "new" natural phenomena. Palm (1995) points out that assessment of probabilities of natural events/hazards are strongly dependent on whether one has experienced an event or not.

Similar to most types of risk analyses RVAfaces the problem of “scope”, and a detailed analysis of every aspect is impossible. Therefore, risk screening is necessary which may follow various axes, e.g., the various project phases, the physical locations of a plant, the mission phases for a flight, and so on. When it comes to flooding, the screening may be difficult. One axis to trace is to consider how the various land areas are linked to waterways. Another axis is to look for various weather conditions, often related to a combination of external influencing factors, such as heavy rain, melting snow from the surroundings, upstream tide, and failure of some of the protective measures. From a methodological point of view it is a challenge to develop a consistent method for screening. Work is ongoing in this field, for example in the project KlimaGIS( where semi-automatic tools are developed to visualize impact of extreme weather conditions on land areas. The basis for such a model is3D maps of the surroundings, and coarse descriptions of the waterways anddrainage systems.

4.3A new method for RVA related to climate change

Based on the implications discussed in the flooding example, it is obvious that extensions of the traditional RVA are needed. These extensions may be integrated into the RVA method for climate change adaptations as follows:

  1. Scope and limitations. The very first step of a RVA is usually to define the scope of the study andlimitations. It is of uttermostimportanceto clarify which decisions the risk analyses shall support. Sometimes the establishment of risk acceptance criteria, trade-offs between risk and economic values (investment cost) are also part of this initial phase.
  2. Screening. The main objective of the screening is to identify relevant land areas for which, e.g., flooding is regarded as an important risk element. The identified spots are often visualized in various geographical information systems in order to get an impression of how issues like how populated the areas are andifindustry is threatened.
  3. Physical response model/drainage analysis. The objective of this step is to establish the necessary physical models describing the flooding situations under various strains (doses). An important part of the modeling is model calibration where calculated values are compared to observed values (in heavy rainfall situations). Another important part of the physical response model is to identify model uncertainty which has not been accounted for in the calibration process. As a first step the impact should be described in terms of which areas are flooded for a given strain. Damage analysis might be treated in a later step for the most critical scenarios.
  4. Dose scenarios identification. The purpose of this step is to identify the various combination of duration/intensity of heavy rain situations, further combined with temperature variation, variation in the snow depth, etc. These scenarios are also deterministic in thesense that for a given identified scenario the impact of the (dose) scenario on the physical response model should be analyzed to reveal the final consequences, e.g., in terms of which areas will beflooded. A huge number of scenarios may be relevant. A challenge is then to select the most representative scenarios. This might require various search algorithms where dose scenarios are propagated through the response models in order to search for similar scenarios but with unanticipated stronger impacts. The probability (or frequency) of each scenario is not a part of the scenario identification, but rough consideration of frequencies is necessary.
  5. Frequency analysis. The purpose of this step is to establish frequencies or probabilities for the various dose scenarios. Frequencies are in the first place established on historical data.
  6. Impact of climatic changes. So-called climatic projections now exist where future scenarios in terms of, e.g. CO2 emissions are taken into account to present some kind of “long term weather forecast”. Such projections exist in Norway on a basis of one by one square kilometerwith respect to precipitation andtemperature (Hanssen-Bauer et.al. 2008). Various models can also predict increase in sea level and other critical factors when it comes to the “dose” modeling. At the time being, the climatic projections are not provided on the format required for the “dose scenario” models, hence post-processing of the climatic projections are required. One of the challenges is that existing climatic projections presents e.g., frequencies of “heavy rain” periods of durations days and weeks, but the most critical part in some of the response models is periods of duration in the order of minutes and hours.
  7. Compiling and presenting risk profiles. The final step in the extended RVA is to combine the deterministic response models with the probabilistic dose scenarios and other uncertainty elements in order to establish a risk profile. When presenting the risk profile, it is important to also present the sources of uncertainty, especially focusing on the “today’s variability”, and the future changes.

Compared to the traditional RVA, step 3-4 and 6-7 are relatively different and require more resources than typically used today. The above-mentioned steps do only include the risk assessment part. Identification of risk reducing measures and emergency preparedness analysis also need to be considered for a complete risk management method, and should be included in, e.g., the preparation of the development plan for a new residential area. Today, these plans often take into account historical data on precipitation, but it is a challenge to take future climate scenarios into account. A challenging question is how much redundancy to put into the drainage system in order to cope with future climatic changes concerned with large uncertainties.