Oil spill modeling

I Introduction

Oil spills have profound impacts on ecosystems (Peterson et al., 2003), the environment (Kingston, 2002), public health (Lyons et al., 1999), the economy (Surís-Regueiro et al., 2007) and communities (Picou et al., 1992). Spills manifest in different ways, both natural and anthropogenic, but in marine environments there are three primary sources: (a) natural seepage (e.g., Coal Oil Point, Santa Barbara); (b) wellhead blowouts (e.g., Deepwater Horizon); and (c) tanker spills (e.g., Exxon Valdez). Although natural seepage rarely yields enough oil to create significant damage to a marine environment, ¹ blowouts and tanker spills can be catastrophic. For example, although the official spill estimates were contested by British Petroleum (BP), approximately 4.9 million barrels (∼205 million gallons) of oil were ejected from the wellhead during the Deepwater Horizon spill in 2010 (Graham et al., 2011). To put this in perspective, 11 million gallons of oil was spilled by the Exxon Valdez in 1989, the largest tanker spill in the U.S. to date (Curl et al., 1992).

These spills, and others like them, have led to a persistent and growing interest to better understand the potential ecological, environmental, social, economic, cultural, and epidemiological implications of catastrophic spills, regardless of the form that they take. In particular, research devoted to quantifying risks, developing models for assessing vulnerability and determining the impact potential of future oil spills remains critical for crafting appropriate response and mitigation strategies. In general, these models and approaches can be characterized by the type of data used and associated outcomes of the study. For example, many “spill specific” studies are conducted ex post, relying on real-world observations and measurements taken directly from the impacted areas. Conversely, ex ante approaches are based on modeled simulations to estimate the final fate of spilled oil and the degree of impact. In addition to the fundamental differences in ex post and ex ante approaches, the underlying methods for spill impact evaluation are also somewhat divergent, especially in data requirements and the quantification of impacts.

The purpose of this paper is to provide a review and synthesis of oil spill risk assessments and impact modeling approaches, as well as their importance for evaluating the spatial vulnerability of communities in/around marine environments. We necessarily adopt a broad definition of community for this paper, where it refers to both human settlements along the coast, but also to larger geophysical and ecological systems where multiple biotic, bacterial, and microbial communities intertwine (Hu et al., 2011; Kennish, 1996; Kostka et al., 2011; Redmond and Valentine, 2012). In particular, our focus on marine spills is important for several reasons. First, the offshore production of oil lease condensate and hydrocarbon gas liquids is growing, accounting for nearly 30% of total global oil production in 2015 (USEIA, 2016). Second, the exploration and production of oil in the marine environments remains both complex and high-risk as drilling operations begin to explore global reserves in deep and ultra-deep water locations (Rocha et al., 2003; Skogdalen and Vinnem, 2012). The stochastic nature of marine environments adds to this complexity. Third, both near shore and open water spills are more difficult to contain, exacerbating the vulnerabilities of proximal communities and ecosystems. Fourth, anthropogenic oil disasters in the marine environment are widely recognized as the most damaging type of spills (NOAA, 2017b), necessitating a more complete evaluation of potential impacts and the development of optimal mitigation strategies.

We recognize that the literature on oil spills, spill modeling, and associated geophysical, environmental, and ecological impacts is both large and dynamic. It is continuously advancing and incorporating new knowledge as it is discovered. Therefore, this review is meant to provide readers and interested scholars with a review of how oil spill impact modeling has been previously explored. In particular we highlight the individual components necessary and the ways of combining the components into a comprehensive risk and/or impact model. Because this progress report centers on spill impact, risk, and spatial vulnerability, we draw liberally from this diverse corpus of knowledge but with a more focused goal of exploring the geographic processes associated with spills and their implications for proximal communities. In the next section, we explore the concepts of risk and vulnerability for oil spills, making an effort to clarify both vocabulary and context. This is followed by a succinct summary of oil spill modeling and its data requirements. Section IV provides an overview of several spill modeling packages that help identify community risk and vulnerability for oiling and details the many ways in which spatial vulnerability manifests both during and after a spill event. We conclude the paper by discussing the comprehensive oil spill impact assessments that bring together each of the component parts.

II Risk and vulnerability

The primary goal of oil spill risk evaluations and impact assessments is to characterize and quantitatively estimate the amount of potential harm that a spill may generate for a particular location. As noted earlier, continued advancements in oil extraction technologies, combined with the push to explore more difficult offshore environments, can generate elevated levels of risk and vulnerability in proximal communities. For example, the threats posed to marine resources from potential oil spills can be enormous, especially in sensitive coastal environments where a complex web of social, ecological, and environmental resources interact (Campagna et al., 2011; Felder, 2009). In the event of a spill, knowing where oil is likely to go, the resources that it may affect, the degree of harm and the overall susceptibility of communities to damage can provide invaluable information for planning, clean-up, and response efforts (Nelson et al., 2015). As more areas are considered for oil extraction, risk and impact assessments are an important step in preventing and mitigating the deleterious effects associated with spills.

III Oil spill models and data requirements

As detailed in the introduction, there are two basic types of oil spill models that are necessary requirements for conducting a related impact or risk assessment, ex post and ex ante. Models that rely upon real-world observations and measurements taken directly from areas impacted by a spill are classified as ex post. Ex ante approaches are based on modeled simulations to estimate the potential location of spills and their degree of impact. A wide variety of data is required for both types of approaches, and there is some data overlap. For example, the most important external data requirement for oil spill models are the data describing the currents and tides within the study area. Most spill models rely heavily on forecast or hindcast oceanographic currents and tides (De Dominicis et al., 2013; Sim et al., 2015: 44), while others use climatological averages of the underlying velocity and direction of currents (Beegle-Krause, 2001). Real-time current and tide data have also been used, but with a simulated level of variation injected into them (Lee and Jung, 2015). In the United States, the Intra-America Seas Nowcast Forecast System (IASNSF), the American Seas Navy Coastal Ocean Model (AmSEAS NCOM) and the Hybrid Coordinate Ocean Model (HYCOM) are freely available oceanographic models that provide researchers with ambient data on ocean conditions at dozens of different depth levels. Wind data, including direction and velocity, are also required for modeling due to the fact that wind is one of the most important aspects for determining surface oil drift and stokes drift (Proctor et al., 1994), and is also used for determining evaporation rates (Geng et al., 2016). Meteorological models, ocean buoy data and/or user-defined model constants are additional sources of wind data if it is not provided in the oceanographic data files.

An additional consideration for accurately simulating an oil spill is understanding the type of hydrocarbon that is spilled, which is contingent upon the location and underlying reserves associated with the extraction area. For crude oils, many assay databases exist to provide researchers with information on their chemical composition. For instance, ExxonMobile provides dozens of crude specific assay files for reference (ExxonMobile, 2017) and the National Oceanic and Atmospheric Association (NOAA) has recently published an extensive online library of different oil types and their properties (NOAA, 2017). Model inputs are not limited to crude oils, in fact many other oil-based products such as processed fuels can also be included. Aside from this baseline information, ex ante oil spill models require data on spill location, duration, depth, amounts, the flowrate (if modeling a blowout) and coefficients (or solutions) for horizontal and vertical diffusion. Much of this same data is required for ex post models. Where the latter is concerned, researchers also have the added benefit of using satellite (Brekke and Solberg, 2005) or drone imagery (Allen and Walsh, 2008) to validate model predictions while systematically tweaking the model to fit the observed empirical results.

Whether the analysis takes place ex ante or ex post, there is a necessary requirement of incorporating quality oil spill data so that the subsequent risk or impact assessment reflects, at least to the greatest extent possible, a form of reality. However, given the geographic diversity of the marine environments and limitations to models in general, there is no single “best” model that would ever be 100% accurate, nor is there any definitive combination of data that should be used for modeling risk and impacts. Context is important, and it is spatiotemporally variable. There is an assortment of spill models that have been designed, and continue to be enhanced as more empirical evidence is gathered on oil behavior in an open ocean system (especially from ex post efforts).

At a minimum, spill models must consider surface and/or subsurface transport, which supplies an indication of where oiling is likely to occur. More advanced models incorporate measures of evaporation, dissolution, entrainment, emulsification, biodegradation, and sediment–oil interaction; all of which enables risk and impact assessment researchers to not only understand the final fate of oil, but also to measure the impact severity (Spaulding, 2017). There are various sources for a more detailed overview of oil spill modeling approaches (Fingas, 1995; Reed et al., 1999; Spaulding, 1988, 2017) and many reviews and comparisons of distinct features built into the spill models (Huang, 1983; Lehr, 2001; Socolofsky et al., 2015). Given the quality and coverage of this material, the nuances of these models will not be covered here. Instead, this review will focus on the more commonly utilized oil spill models (detailed in the next section). The outputs of these models are often used for risk and impact analysis and we attempt to detail the ways in which this is achieved.

IV Modeling spill risk, vulnerability, and impact

Modeling oil spill risk, vulnerability, and impact requires a range of data and tools. Risk needs to be calculated to establish the likelihood of a spill event and its spatiotemporal footprint, and the spatial vulnerability of communities needs to be evaluated in order to capture and quantify the susceptibility of a location to the harmful effects of oiling. Impact assessments represent a coupled composite of risk and vulnerability, where the potential outcomes of oiling are evaluated in formal economic, ecological, cultural, political, geographical, and environmental terms. When placed into a logical framework, this combination of measures can represent the magnitude of each effect (individually) and the total effect of oil overall. The result is a more holistic approach to understanding the consequences of an oil spill event—one that reflects important aspects of what actually occurs during and/or after a spill event.

4.1 Modeling risk

Modeling oil spill risk can be accomplished in several different ways, but the overarching goal is to determine the likelihood of oil occurrence and the degree of oiling an area might experience (Gasparotti, 2010). The primary tool for modeling risk is numerical spill simulation (Amstutz and Samuels, 1984), which can be (but not always) informed by data on historical spill events (Fernández-Macho, 2016). Risk is, in general, conceptualized as the probability of an event occurring with respect to some number of previous events. Formally

r i s k ( r ) = n ( A ) n

1

where there is some number of events n that occurred at/in (A) (which could be particle landfall in a specific location) divided by the total number of possible events n (particles released). This is most commonly derived through simulations (informed by numerical models), which are proving to be some of the most powerful and popular methods used for determining risk based on probability of occurrence (see equation (3)). In part, this is due to the relative growth, development, accuracy and access to spill simulation packages. Some of the more commonly used models are GNOME (Beegle-Krause, 2001), MEDSLICK (De Dominicis et al., 2013), SIMAP (McCay, 2003), and OSCAR (Aamo et al., 1997), but there are many others. Readers should consult Spaulding (2017) for a thorough review of current oil spill modeling tools. Many researchers are also developing spill and transport models outside of the more common “packaged” oil spill modeling platforms. For example, Lee and Jung (2015) simulate oil behavior using a Lagrangian model to track individual oil particles, which is coupled with their hydrodynamic model for tidal and wind forcing, and NOAA’s ADIOS tool to determine weathering effects. In other work, Naidu et al. (2013) develop a model to track and estimate the final fate of spills using a boundary-fitted grid technique, which can then be used to develop risk metrics.

Following the simulation of an oil spill, calculations of risk can take place in a number of ways. The use of ensemble approaches, where hundreds or thousands of spills are simulated at multiple locations, is an increasingly popular option (Sepp Neves et al., 2016). For example, Boer et al. (2014) simulated 3500 spills in a coastal lagoon in Portugal with a lesser known spill model. The end product of this effort was a suite of hazard maps based on probability of oil occurrence. Similarly, Guillen, Rainey and Morin (2004) use the former Mineral Management Service’s Oil Spill Risk Assessment Model (OSRAM) to calculate the probabilities of coastal oiling by simulating 2018 spills at each of 91 different launch points in the Gulf of Mexico, over a simulated nine-year period. Al Shami et al. (2017) uses MEDSLICK to simulate 360 spills to calculate the probability of oiling and mean amount of beached oil in the area of interest. Finally, NOAA’s Trajectory Analysis Planner (TAP) utilizes GNOME outputs to determine a probability of occurrence and likely impacts based on the spatial distribution of the generated oil parcels (Barker, 1999).

In addition to the one-off simulations and ensemble approaches, alternative strategies to determine risk have been developed. For example, Monte Carlo approaches are used to generate risk distributions based on deterministic and uncertain information (Li et al., 2014) associated with spills, as well as to determine oil spill risk distributions based off a representative sample of spills in a particular location (Nelson and Grubesic, 2017). Risk can also be quantified based on proxies of where oiling is likely to occur given the location of oil-related infrastructure. For example, Mokhtari et al. (2015) use a generalized linear model to estimate probabilities of oiling based on ship density along designates routes, coastlines, oil facilities, and oil well locations in the Persian Gulf. Historical spill incidents have also been used to determine the risk level posed by oil spills to European coastlines. Specifically, Fernández-Macho (2016) uses a database of 10,000 spills emanating from oil carriers and barges since 1970 to estimate risk.

4.2 Modeling vulnerability

Following on from the three forms of vulnerability that were detailed previously, it is important to acknowledge that a complimentary definition exists in the context of oil spills. Specifically, Dow (1999) defines vulnerability as the differential susceptibility of ecosystems, households, or social groups to losses, and is a function of the level of exposure to the risk, ability to withstand impacts, and the ability to recover from impacts. This definition is very close to the idea of spatial vulnerability detailed earlier, combining elements of the physical and social, but it neglects the explicitly spatial characteristics of vulnerability required for a comprehensive evaluation of potential impacts. Vulnerability calculations generally try to capture the variation in spatial characteristics of an area. A general form of vulnerability can be represented as

v u l n e r a b i l i t y ( v ) = ( ∑ i = 1 n S 1 i + S 2 i + S 3 i +…+ S n i ) ( o i l k o i l max )

2

where vulnerability is the sum of each sector score (S₁, S₂, S_n) within each individual unit of analysis (i), keeping in mind that sectors can be environment, social, economic or others. This is then multiplied by the severity of oiling, which is normalized by some value. The normalization procedure scales values for equation (2) to reflect both vulnerability and severity of oil. In this subsection, we review the various ways in which community vulnerability is captured for oil spills, emphasizing its spatiotemporal facets when possible.

4.21 Environmental sensitivity

One of the most important elements of oil spill vulnerability measurement and assessment is accomplished through the use of an ESI. Initial development efforts associated with the ESI were focused on determining shoreline interaction with the physical processes controlling oil–land interaction and converting this interaction into a scale that ranged from 1–10. Here, ESI values reflected the level of oil deposition, observed prevalence and longevity of oil in the environment, as well as the extent of biologic damage (Gundlach and Hayes, 1978). For example, exposed, rocky outcrops are assigned a value of 1 because the wave action and lack of permeability keeps most of the oil offshore. Coarse-grained sand beaches are assigned a value of 4 because oil may infiltrate the surface, but in high energy conditions oil may be naturally removed within a few months. Salt marshes and other highly vegetative locales receive a 10, not only because they are the most productive aquatic environments, but because they are also easily harmed. In these locations, oil can persist for years and clean-up is difficult (Gundlach and Hayes, 1978).

Today, the two most common methods for ESI classification are field observation (with subsequent digitization) (Grundlach and Hayes, 1978) and remote sensing (Carmona et al., 2012; Jensen et al., 1990). In many cases, both approaches are used simultaneously, but increasingly remote sensing is the preferred method. Subsequent development of the ESI have included enhancements to incorporate other shoreline habitats and human-use areas (Jensen et al., 1998), improved classification algorithms for imagery (Schiller et al., 2005) and field work to better determine acute levels of harm susceptibility (Scott et al., 2013). NOAA authors an extensive ESI database and provides guidelines for classification type (NOAA, 2017a). One major limitation of ESI is that although human use and biological resources are identified in the index, only the shoreline vulnerability is classified (Santos and Andrade, 2009), largely neglecting proximal areas and communities.

Major limitation aside, the ESI classification process is both popular and widely used, with researchers modifying or creating additional ESI codes when necessary (Owens and Robilliard, 1981). For example, field data collection and digitization has been emphasized in many studies, such as Nansingh and Jurawan (1999), who took field observations of the Trinidad coastline and built them into an ESI scale for shoreline segments reflecting the exposure to wave and tidal energy, sediment type, slope, and biological productivity (Table 1). One reason that field observations remain important for evaluating ESI and associated vulnerability is because they can help decrease observational error when only remotely sensed methods are used. Once the field data are collected, researchers integrate, manage, and analyze them with geographic information systems (GIS), frequently using the NOAA classification schemes to assign values to the field data (Adler and Inbar, 2007; Aps et al., 2016; Fattal et al., 2010; Wieczorek et al., 2007). As alluded to earlier, many researchers are also using hybrid approaches for ESI development and vulnerability assessment by conducting field observations and interviews with experts, and collecting satellite data. For example, the work by Abdel-Kader et al. (1998) focused on the vulnerability of Egyptian shorelines, using SPOT satellite imagery, existing research from the area, personal communication with specialists, and local inhabitants to map the shoreline ESI. Pincinato et al. (2009) used secondary surveys, remote sensing, and field studies to build a cartographic database of ESI classifications for shorelines in Brazil.

OPEN IN VIEWER

Table 1.

NOAA’s Environmental Sensitivity Index classifications reflecting the susceptibility to damage by oiling (NOAA, 2017a).

Shoreline type	ESI index value
Exposed rocky shores	1
Exposed rocky platforms	2
Fine-grained sand beaches	3
Coarse-grained sand beaches	4
Mixed sand and gravel beaches	5
Gravel beaches	6A
Riprap structures	6B
Exposed tidal flats	7
Sheltered rocky shores	8A
Sheltered man-made structures	8B
Sheltered tidal flats	9
Salt to brackish marshes	10A
Freshwater marshes	10B
Swamps	10C
Mangroves	10D

4.22 Environmental vulnerability

Although raw ESI values provide a good indication of the potential harm a shoreline will experience if exposed to oil, this information requires a transformation, to index form, to provide stakeholders with an easily-interpreted indicator of environmental damage susceptibility. In addition, it is important to acknowledge that neither wildlife nor biological resources are necessarily incorporated into the ESI value for a particular shoreline. Instead, values for these resources are often binary in nature, reflecting the presence or absence of a species in an area. It is also important to acknowledge that ESI values are not required to establish varying levels of environmental vulnerability. Indeed, many studies have used alternative methods to calculate environmental vulnerability, which is usually described as the combination of physical (surface type) and biological vulnerability.

Consider, for example, the development of a physical vulnerability index value. A common approach demonstrated by Castanedo et al. (2009) is to explicitly consider the physical coastline factors in a manner that reflects the original ESI classifications, but then add supplementary measures such as shoreline exposure to wave energy, sinuosity, and coastal slope. These additional measures allow researchers to determine a “self-cleaning” rate. In other words, it reflects the degree to which oil will naturally be washed away through wave action. Such measures can be combined to give a rating of exposure (e.g., least/most exposure to wave energy) that is arithmetically added to the slope score of a shoreline. The resulting value is a total physical vulnerability index. The same broad measurements of physical vulnerability are used by many, including Cai et al. (2015), Canu et al. (2015), Fattal et al. (2010), and Mendoza-Cantú et al. (2011). In addition to the physical vulnerability of a shoreline, it is also possible to develop and add a permeability index while assigning weights based on ESI classifications. In fact, permeability is a highly important measure that further reflects the rate that a shoreline can naturally rid itself of oil (Moe et al., 2000).

The development of biological vulnerability index is a more complex and nuanced process when compared to the development of physical vulnerability indices. Where the latter is concerned, ESI classifications assign vulnerability scores based on geomorphological structure alone (Santos and Andrade, 2009). For biological vulnerability indices, common practice is to map and subsequently incorporate known, ecologically sensitive areas. Assignment of index values is based on the regulatory protection (federal, state, local) and/or a ratio of total sensitive shoreline length to overall shoreline length (Cai et al., 2015; Frazão Santos et al., 2013). If data allow, biomass, biodiversity, and habitat type can be quantified and incorporated as sub-indices, describing the larger biological category (Fattal et al., 2010; French-McCay, 2004; Lan et al., 2015). It is also common for studies to take a more generalized approach by collapsing multiple bio-indicators into one category and assigning a vulnerability value (Romero et al., 2013). Another example of a more generalized approach is taken by de Andrade et al. (2010), who construct a “natural” vulnerability index with a summation of values ranging from 1 to 3. These scores reflect the combination of flora sensitivity (e.g., mangrove areas, beach and coastal plateau) and the type of land use (e.g., fishing, leisure and habitancy). Ultimately, the vast majority of biological vulnerability indices reflect the availability of data for an area and local knowledge/subjective assessment by researchers as to what best captures the vulnerability of biologic resources in a study area.

4.23 Social and economic vulnerability

Social and economic vulnerability indices attempt to capture the ability of coastal residents, communities and economic systems to absorb and recover from the damage caused by an extreme oiling event. This is particularly salient for communities that have a measurable dependence on the marine environment for their well-being. Over the past two decades, several large-scale data collection efforts have allowed for a more robust assessment of social and economic vulnerability (Cooper and McLaughlin, 1998), but many of the most current efforts to capture social vulnerability fail to include anything other than population size (e.g., Frazão Santos et al., 2013; Sepp Neves et al., 2015). In other work, measures associated with population marginalization (Mendoza-Cantú et al., 2011) and residential land use (Liu et al., 2015) are accounted for, as are elements of the mental health of community residents after a spill (Allen and D’Elia, 2015; Grattan et al., 2011; Lazarus, 2016). However, as mentioned previously, a more detailed treatment of social vulnerability should include race, income, occupation, age, ethnicity, and other important measures that can reflect the diversity of a community and the varied experiences of its residents after a major spill. For best practices in this regard, although not necessarily related to oil spills, see Cutter and Emrich (2006), Cutter et al. (2003), Tate (2013) and Wu et al. (2002).

In much of the oil spill literature, there tends to be a larger emphasis on measuring the economic vulnerability of an area. Data used to measure economic vulnerability reflects the level of dependence that a particular activity places on the marine environment, which can include marine transportation, tourism, recreation, and the natural resource industries (Azevedo et al., 2017; Cai et al., 2015; Castanedo et al., 2009; Wirtz and Liu, 2006). Quantification of the vulnerability can be achieved by estimating the total number of vulnerable assets in a location and assigning an index value (Nelson et al., 2015) or the estimated income lost for each of the economic sectors effected by a spill (Wirtz et al., 2007). Regardless of the variables used, “sectors” or “bins” that capture the number of activities in a particular area are helpful for determining an index value to represent the differing levels of potential harm caused by an interaction with oil (Fattal et al., 2010; Sepp Neves et al., 2015). In the instances where economic vulnerability is dissected into multiple sub-indices, those indices are combined to represent the overall economic vulnerability to oiling.

4.24 Global vulnerability and/or spatial vulnerability

Once the subsets of vulnerability indices are created for specific sectors (e.g., physical, social, biological), researchers can combine them to create a global measure of vulnerability for each unit of analysis. If spatial data are explicitly tied to the index values, this yields a comprehensive index of spatial vulnerability. If the measures include aspatial elements, the result is more akin to a global vulnerability measure (Birkmann, 2006). Regardless of the specific context or data used, a simple summation of index values across sectors and/or weights can be applied (Bauer et al., 2015: 60; Cai et al., 2015; Kankara et al., 2016). When weights are used, a normalization procedure is required, prior to weighting, to ensure that each facet of vulnerability exists on a common scale. In doing so, the weight applied to the sectors gives a meaningful change in the overall vulnerability relative to the other sectors (Fernández-Macho, 2016; Mokhtari et al., 2015). A Min-Max procedure is commonly used to normalize all data to an identical interval (Fattal et al., 2010; Frazão Santos et al., 2013) or values can be divided by the value range to give an interval from 0 to 1 (Castanedo et al., 2009; Wirtz and Liu, 2006). Other methods will take continuous data and divide it into bins using a classification method such as Jenks natural breaks or equal interval (Jenks and Caspall, 1971; Nelson et al., 2015). After normalization, weights are multiplied by the vulnerability level to generate a global and/or spatial vulnerability score for each unit in the study area.

V Impact assessment

Once risk and vulnerability have been determined, impact measures can be calculated. As detailed previously, impact is a function of the likelihood of oil occurrence, the severity of the oiling, and the vulnerability level of an area. As each facet increases, the predicted impact will also increase and can be summarized as follows

i m p a c t = f ( v , r )

3

There is, of course, some flexibility in the way that the impact measure can be defined, but impact assessments typically reflect a combination of risk (equation (1)) and vulnerability (equation (2)).

Consider, for example, the general approach to oil spill impact modeling as demonstrated by Kankara et al. (2016). In it, ESI classifications, scientific value, environmental importance, and level of economic activity contribute to a developed risk index (Kankara and Subramanian, 2007), which is then combined with GNOME spill outputs. A complete classification of impact is given by estimated number of species expose, and a protection priority index for coastal segments. The general approach is replicated by Azevedo et al. (2017), with a hazard assessment based on oil transport and exposure and a vulnerability assessment focusing on biological, socio-economic and physical indicators. Final impact is determined using an equation similar to equation (1), where generated values yield a normalized risk index ranging from 0 to 1. In fact, most impact assessments use a similar approach, including the work of Olita et al. (2012) and Canu et al. (2015), who use a normalized index of total impact, Sepp Neves et al. (2015), who combines risk and vulnerability to estimate impact along the Lebanese coast, and Al Shami et al. (2017), who uses a modified ESI and risk probability to determine an overall impact metric for the coast.

Again, it is important to reiterate that oil spill impact modeling may use additional methods and measures that may change depending on data availability and location (Romeo et al., 2015). For example, one of the most complete and highly cited impact models is detailed in French-McCay et al. (2004). Spills are simulated using SIMAP and combined with the NOAA ESI values to measure coastline vulnerability. Species abundance measurements are also incorporated to give an overall measure of wildlife injury. Impact, at least in this study, is reported in terms of the National Resource Damage Assessment (NRDA) framework and focuses on the ecological costs of restoration, accounts for the type of hydrocarbon (gasoline, diesel, crude oil, and heavy fuel oil), and level of predicted damage. The work builds on previous research found in McCay (2003), French-McCay (2004), and Etkin (1999). GNOME has been used in a similar manner to estimate the degree of oiling, measured as total hydrocarbon concentration, and compared against species sensitivity distributions to calculate the proportion of species in an area that would be affected by the oil under varying dispersant applications (Bejarano and Mearns, 2015). Bayesian networks have also been used to estimate the difference in the probability distributions of acute impacts of an oil spill on a variety of organisms in the Gulf of Finland based on a “most probable accident” and a “worst case scenario” (Lecklin et al., 2011).

As oil exploration moves into new offshore frontiers, there will be a need to adapt existing data and methods used for impact assessment to those regions. Romeo et al. (2015) provide a useful approach for comparing two regions and assessing where improvements or modifications will need to be made. Their process of a systematic comparison between two differing regions offers a useful framework for future researchers wishing to extend existing methods and techniques to new areas.

VI Conclusion

This paper has presented the necessary components for providing a thorough, yet concise, overview of oil spill risk, vulnerability, and impact modeling as they have been presented and developed within the literature. We hope that this progress report provides researchers with a sensible starting point when undertaking studies in this domain. The most important part of any impact assessment is the accuracy of data and the real-world representativeness of likely damage. Weights and predictions associated with damage can be highly subjective. Thus, it is always best to validate the index values used with real-world results. Because oil exploration and extraction continue to be large contributors to the global oil market, these analyses will remain useful and important. It is wise, therefore, to continue to develop and enhance these models; taking note of the fundamental components developed over several decades and adding new data, methods and techniques as they are made available to the research community.