6.2.2 Risk assessment methodologies

A number of different assessment methodologies have been and are being applied to CCS-related projects. The main methodologies being used are: (a) scenario analysis, analyzing how a CO2 storage system might evolve in particular in terms of CO2 migration/leakage, (b) fault / event tree analysis, to evaluate as a combination of possible steps the network of pathways for CO2 release and migration starting from thestorage reservoir and ending at a particular point of interest, (c) expert judgment, to derive from relevant experience and expertise in a specific area, the likelihood of CO2 leakage, (d) screening-level analysis, which can be useful to compare safety characteristics of different sites based on expert opinion (Stenhouse et al., 2009). In parallel with methodologies, a variety of approaches are available for mathematical modecoing, which can be classified under three general categories: (a) numerical models, which use discretization methods to model detailed processes describing the system evolution over space and time, (b) analytical / semi-analytical models, which are mathematical models in which the solution to the equations used to describe changes in the system can be expressed as an analytical or semi-analytical function, typically as a function of time in the case of risk for CCS projects, and (c) compartment or mixing-cell models, representing a large family of models, where the model comprises a series of individual compartments representing different physical domains of the total storage system. All of the above-mentioned models can be run deterministically or probabilistically (Stenhouse et al., 2009).

Risk methodologies are generally classified in two main groups: qualitative and quantitative.

  • Qualitative Risk does not provide concrete or numerical results. In case of a lack of data and/or specific knowledge, time and expertise, qualitative risk may be sufficient and more effective. Among the most common qualitative methods are the Features, Events, and Processes (FEP), and the Vulnerability Evaluation Framework (VEF).

  • Quantitative Methods are used in well-known systems where the level of uncertainty is relatively low. Two main kinds of methods belong to this group: Deterministic Risk (DRA) and Probabilistic Risk (PRA). DRA does not handle uncertainty, but is useful in determining trends due to its single parameter variation. It gives very accurate results when the input parameters are exactly known. PRA, on the other hand, can statistically quantify the uncertainty associated with parameters describing the processes in deterministic models. PRA is the most preferable method of assessing long-term risk in complex systems (Condor et al., 2011).

Table 6.1 summarizes the main characteristics of some risk methods. It should be noted that some methods are not considered in this table, e.g. Preliminary Hazard Analysis (PHA), Failure Mode and Effects Analysis (FMEA), Fault Tree Analysis (FTA) and Fuzzy Logic.

E. Tab . 6-1

Tab. 6-1: Risk Methods (Condor et al., 2011).

The Features, Events and Processes (FEP) method consists of listing relevant factors that describe the current state and possible future evolution of a site. The FEP analysis is useful in the licensing and certification stages of project development..

The Vulnerability Evaluation Framework (VEF) is a qualitative method which systematically identifies conditions that could increase or decrease the potential for adverse impacts (i.e., susceptibility to consequences). The VEF (EPA, 2008) is not designed as a site selection tool, it does not aim to establish performance standards, or to specify data requirements. It is a conceptual framework designed to help regulators and technical experts in framing specific considerations and identifying areas that require design evaluation, specific risk, monitoring, and management (Condor et al., 2011).

The Risk Identification and Strategy using Quantitative Evaluation (RISQUE) proposed by Bowden and Rigg, 2004 is a systematic quantitative process based on the judgment of a panel of experts. It delivers a transparent risk in a process that can interface with the wider community and allow stakeholders to assess whether the CO2 injection process is safe, measurable and verifiable. It has been applied in Australia, under the GEODISC research program, to assess the risk posed by conceptual CO2 injection in four selected areas (Dongara, Petrel, Gippsland, and Carnavarcon). The approach relies on quantitative techniques to characterize Risks in terms of both likelihood of identified risk events occurring (such as CO2 escape and inadequate injectivity into the storage site) and consequences (such as environmental damage and loss of life). It consists in five stages:

Stage 1- Establishing the context, i.e., assessment of the nature of the activities and potential impacts,

Stage 2- Risk identification,

Stage 3 - Risk analysis, i.e., quantification and modeling of probabilities and COnsequences for each substantive risk event,

Stage 4 - Development of risk management strategy, i.e., defining and evaluating options for action plans to treat key risk events,

Stage 5 - Implementation of the risk management strategy (Bowden and Rigg, 2004).

RISQUE methodology in conjunction with a modified Delphi approach was proposed for assessing and quantifying risk in CO2 geological storage projects aiming at the reduction of uncertainties. The RISQUE process does not routinely include a continual and progressive technology risk component but the modified Delphi process could address this potential need. The RISQUE method addresses the risks but in a linear scheme, whereas in complex programmatic settings, risks are considered in a non-linear scheme, chosen by interveners or stakeholders. The modified Delphi technique, when planned and implemented well, can bring all elements of risk, presumably from the non-linear 'risk-space' into a controlled input for the RISQUE process. It quantifies and qualifies the risks perceived by others into a set of consensus risks, in a re-iterative agreement process, and weight those factors against the expertpanel of the RISQUE process. As the RISQUE process proceeds, the Delphi re-iterative process continues using smaller subgroups and continued quantified and qualified weighted input (Wyatt et al., 2009).

The Structured What-If Technique (SWIFT) is a form of Delphi risk analysis used for qualitative hazard identification that was attempted by Vendrig et al., 2003, who identified major hazards through a "Structured What-If Technique" involving an expert panel (CSLF, 2009). The method was developed as an efficient alternative to the Hazard and Operability (HAZOP) technique and to the Failure Modes and Effects Analysis (FMEA) for providing highly effective hazard identification in situations and systems where none of them were adequate. It consists of a series of "what-if…?" or "How could…?" questions to identify situations, issues or threats of potential harm. There is no single standard approach to SWIFT which is flexible and has to be modified to suit each individual application (Vendrig et al., 2003; Condor et al., 2011).

The Certification Framework (CF or CFA) is a simple risk approach for evaluating CO2 and brine leakage risk at GCS sites (Oldenburg et al., 2009). It is similar to VEF, but it adds values for the leakage probability (Condor et al., 2011). Its purpose is to provide a framework for project proponents, regulators, and the public to analyse the risks of geologic CO2 storage in a simple and transparent way to certify start-up and decommissioning of storage sites. The CF currently emphasizes leakage risk associated with subsurface processes and excludes compression, transportation, and injection-well leakage risks. It is designed to be simple by using (a) proxy concentrations or fluxes for quantifying impact rather than complicated exposure functions, (b) list of pre-computed CO2 injection results, and (c) simple framework for calculating leakage risk. For quantification of risk, the system is divided into compartments that can be subsurface (hydrocarbon reservoirs or underground sources of water), at surface (local sites where leakage occurs) and distant sites (Condor et al., 2011). The CF approach has to be based on a clear and precise terminology in order to communicate to the full spectrum of stakeholders:

  • Effective trapping (proposed overarching requirement for safety and effectiveness),

  • Storage region (3D volume of the subsurface intended to contain injected CO2),

  • Leakage,

  • Compartment (region containing vulnerable entities, e.g. environment and resources),

  • Impact,

  • Risk,

  • CO2 leakage risk (risk to compartments arising from CO2 migration, i.e. the product of the probability of intersection of leakage paths with compartments (Oldenburg et al., 2009; Dodds et al., 2011).

In the CF, impacts occur to compartments such as HMR (Hydrocarbon and Mineral Resource), HS (Health and Safety), USDW (Underground Source of Drinking Water), NSE (Near-Surface Environment), and ECA (Emission Credits and Atmosphere). Wells and faults are assumed to be the only potential leakage conduits. Fig. 6-4 shows the CF conceptualization of the system into source, conduits and compartments (left-hand side), and a flow chart of the general CF logic and inputs and output (right-hand side). A similar method to the CF approach is the Screening and Ranking Framework (SRF) which is based on the assumption that if the primary seal or containment leaks, the second seal will act. If the second seal fails, then the leakage will be attenuated or dispersed (Oldenburg, 2008; Condor et al., 2011).

E. Fig . 6-4

Fig. 6-4: Generic schematic of compartments and conduits in the CF (left-hand side), and flow chart of the CF approach (right-hand side) (Oldenburg et al., 2009).

The Multi-Criteria Assessment (MCA) covers a variety of non-monetary evaluation techniques sharing a basic framework under which a number of alternatives can be scored against a series of defined or fixed criteria. This list of criteria is proposed according to the fundamental goals of the CGS. These criteria can then be categorized in groups. Multi-criteria assessment (MCA) appears to hold much potential as a useful tool for characterising and better understanding differences in stakeholder assessments of CCS and its implications, and for identifying options around which greater consensus on the desirability (or otherwise) of CCS as a mitigation strategy might emerge (Gough and Shackley, 2006). This method delivers a rich profile of the views and preferences of participants and thus enables 'mapping' key issues that will affect the prospects for further development. A similar method is the Multi-Attribute Utility Theory (MAUT). The main difference between MAUT and MCA is that MAUT assumes a dependency of preferences of criteria, enabling the inclusion of subjective elements (Scholz and Tietje, 2002; Condor et al., 2011).

The Evidence Support Logic (ESL) has been designed to identify the amount of uncertainty or conflict involved in a decision. This involves systematically breaking down the question under consideration into a logical hypothetic model whose elements expose basic judgements and opinions related to the quality of evidence associated with a particular interpretation or proposition. A decision-support tool called TESLA implements Evidence Support Logic (ESL). The method involves constructing decision trees to reflect: (1) the Performance Assessment's context since decision depends on the storage project's stage of development and the aims of the stakeholders); (2) the FEPs that may influence the system being evaluated; (3) the kinds of information that enable assessments about the characteristics and effects of interactions among these FEPs. The decision tree consists of a hierarchy of hypotheses, which links the main hypothesis of interest (e.g. insignificant CO2 leakage from a deep storage reservoir) to data or information (e.g. geological evidence for the existence of a cap rock, experimental evidence for the effective sealing of boreholes, output from supporting modelling studies etc). The 'evidence' for or against each hypothesis is the extent to which information leads to confidence in the hypothesis' dependability or falsehood respectively (Metcalfe et al., 2009). The 'evidence' may correspond to quantitative information (e.g. numerical model output, measurements in boreholes etc) or qualitative information (e.g. anecdotal evidence that a particular kind of borehole seal is effective). Each item of qualitative or quantitative information is then mapped into two values on a numerical scale of 0 to 1 representing evidence for and against. This representation of evidence is a type of Interval Probability Theory, which employs three-value logic. Experts assign values to each hypothesis representing the amount of supporting evidence, the amount of refuting evidence and the amount of uncertainty or conflict in the evidence. (Metcalfe et al., 2009; Condor et al., 2011).

The Method Organized for a Systematic Analysis of Risk (MOSAR) allows the analysis of the technical risks of a system and then identifies the prevention means in order to neutralize them. It consists of two main steps (Fig. 6-5). First step, 'A', allows the analysis of major risks. Second step 'B' makes a detailed analysis of project implementation and specifically defines the safety tools related to the technical dysfunction (all dysfunctions are found with this step). The MOSAR method is a systematic method which relies on a step by step method, in which no step can be neglected. This does not prevent flexibility: when an unexpected event arises or a new danger source appears, it can included at the beginning of the method without changing all the process. This method is built level by level where each level gives a specific information so that it is possible to stop at a chosen level. Unexpected events, such as physical harm and material, fauna, flora, ground and economic damages or unpleasant effects on the population, are defined. MOSAR is based on site observations and facts and is applicable to a specific installation because it accounts for technical aspects, site morphology and geology, politics, and economic and social aspects into account. It presents the advantage of creating improbable and unforeseeable risk Scenarios with a first analysis, but whose implementation can be extremely beneficial. The important subjectivity of MOSAR has been noticed and this should be viewed as a strong point and not as a hindrance (Cherkaoui and Lopez, 2009).

E. Fig . 6-5

Fig. 6-5: The MOSAR method: Steps A and B (Cherkaoui and Lopez, 2009).

The Performance and Risk (P&R) assessment (or Performance and Risk Management methodology - P&RTM), for well integrity was developed by Schlumberger and OXAND. The uncertainties of the system are converted into the notion of probability and the quantity of CO2 leakage mass assessed into the notion of severity. It also includes the definition of a Risk Acceptance Limit (RAL), which brings forwards the criteria of unacceptable risks. The methodology is based on experience in material ageing and risk of complex systems, where probabilistic simulations are performed. It accounts for all stakes involved in well integrity management and enables the full integration of uncertainties as part of risk estimation. The methodology improves common approaches based on FEPs as it quantifies risk levels. It provides useful and reliable tools to support decisions for well integrity management strategies or emergency plans. Updating risk with incoming data allows an evolving vision of risk levels to optimize interventions in time. The main objectives of the risk-based methodology regarding well integrity are to identify and quantify risks associated with CO2 leakages along wells over time (from tens to thousands of years), to evaluate risks and to propose relevant actions to reduce unacceptable Risks (Le Guen et al., 2008; Le Guen et al., 2009; Condor et al., 2011)

A hybrid system-process model CO2-PENS (Predicting Engineered Natural Systems) is a probabilistic simulation tool designed to incorporate CO2 injection and sequestration knowledge from the petroleum industry to perform risk of sites. It includes economic tools, as well as models for the physical and chemical interactions of CO2 in a geological reservoir (Viswanathan et al., 2008; Stauffer et al., 2009). This model is based on a PID (Process Influence Diagram)-like approach extending the FEPs analysis. The CO2-PENS tool aims at integrating in a system-level model a number of process-level models representing the storage reservoir, the cap rock, the potential release mechanisms, the transport of CO2 from the reservoir and the release of CO2 in surface.The CO2-PENS system model allows both a simplified analytical description of processes and the use of detailed processmodels. It links high level system models (i.e., reservoir model) to the process level (wellbore leakage, chemical interaction of CO2) and represents thus a hybrid coupled process and system model designed to simulate CO2 pathways, such as capture, transport, injection into geological reservoirs, potential leakage from the reservoir and migration of escaped CO2. Due to its modular architecture, the tool allows incorporation of additional process models by linking to dynamic linked libraries (DLL) and coupling of the well leakage module with the atmospheric model is feasible. At each time step in the system model, the wellbore module is queried to predict the leakage rate into the top aquifer. Simulation of wellbore leakage is complicated and simulation approaches require PDFs with respect to potential failure mechanisms as input parameters to take account of uncertainties. CO2-PENS is being used inrisk assessments for several of the field tests and demonstrations being conducted as part of the United States Department of Energy's (US DOE's) Regional Carbon Sequestration Partnership efforts (CSLF, 2009).

The System Modelling Approach (SMA) is part of the CO2-PENS and was developed in Los Alamos National Laboratory and originally designed to perform probabilistic simulations for the whole CCS chain. The long-term fate of the injected CO2, including possible migration patterns out of the target formation, is simulated through probability distributions (Stauffer et al., 2009). Oldenburg and Bryant, 2007 decompose the system into process-level models. They focus on a simple certification framework. The storage complex is divided into compartments. The likelihood of a leak is evaluated by estimating the probability that a leakage pathway encounters the CO2 plume on the one side, and a target on the other side. The CO2 flux across the pathway is simulated through deterministic simplified models, and the impacts of the release compared to acceptable thresholds. A level of risk is obtained by the product of the values of the probability and the consequences (CSLF, 2009).

Researchers for the Weyburn CO2 Monitoring and Storage Project have developed a program called CQUESTRA (CQ-1) and applied it to components of the project (Whittaker et al., 2004). The probabilistic conceptual model (PCM) consists of two components: the model domain, which defines the geologic setting, and the model processes, which include the physical and chemical processes that define CO2 mass transport and storage. The model domain is divided into four broad areas: the biosphere, the upper geosphere (all aquifers and aquitards above the reservoir), the wells, and the lower geosphere (reservoir and the aquifers and aquitards below the cap rock). Local variability in formation porosity, permeability, Darcy flow velocity, etc., is incorporated into Probability Distribution Functions (PDFs) to capture the uncertainty in the PCM's domain features and processes. Once the physical PCM domain is fully described, CQ-1 quantifies the main driving forces relevant to the storage of CO2 in a reservoir. CQ-1 was used to model the Weyburn system for a period of 5,000 years after completion of EOR CO2 injection. A Monte Carlo simulation method was used to sample the probability distribution functions for the CQ-1 input parameters (Deel et al., 2007).