- Top of page
- INDIVIDUAL-BASED MODELS
- RESULTS AND CONCLUSIONS
Ecological risk assessments need to advance beyond evaluating risks to individuals that are largely based on toxicity studies conducted on a few species under laboratory conditions, to assessing population-level risks to the environment, including considerations of variability and uncertainty. Two individual-based models (IBMs), recently developed to assess current risks to sea otters and seaducks in Prince William Sound more than 2 decades after the Exxon Valdez oil spill (EVOS), are used to explore population-level risks. In each case, the models had previously shown that there were essentially no remaining risks to individuals from polycyclic aromatic hydrocarbons (PAHs) derived from the EVOS. New sensitivity analyses are reported here in which hypothetical environmental exposures to PAHs were heuristically increased until assimilated doses reached toxicity reference values (TRVs) derived at the no-observed-adverse-effects and lowest-observed-adverse-effects levels (NOAEL and LOAEL, respectively). For the sea otters, this was accomplished by artificially increasing the number of sea otter pits that would intersect remaining patches of subsurface oil residues by orders of magnitude over actual estimated rates. Similarly, in the seaduck assessment, the PAH concentrations in the constituents of diet, sediments, and seawater were increased in proportion to their relative contributions to the assimilated doses by orders of magnitude over measured environmental concentrations, to reach the NOAEL and LOAEL thresholds. The stochastic IBMs simulated millions of individuals. From these outputs, frequency distributions were derived of assimilated doses for populations of 500 000 sea otters or seaducks in each of 7 or 8 classes, respectively. Doses to several selected quantiles were analyzed, ranging from the 1-in-1000th most-exposed individuals (99.9% quantile) to the median-exposed individuals (50% quantile). The resulting families of quantile curves provide the basis for characterizing the environmental thresholds below which no population-level effects could be detected and above which population-level effects would be expected to become manifest. This approach provides risk managers an enhanced understanding of the risks to populations under various conditions and assumptions, whether under hypothetically increased exposure regimes, as demonstrated here, or in situations in which actual exposures are near toxic effects levels. This study shows that individual-based models are especially amenable and appropriate for conducting population-level risk assessments, and that they can readily be used to answer questions about the risks to individuals and populations across a variety of exposure conditions. Integr Environ Assess Manag 2012; 8: 503–522. © 2012 SETAC
- Top of page
- INDIVIDUAL-BASED MODELS
- RESULTS AND CONCLUSIONS
Ecological risks and ecological recovery following environmental disturbances are critical issues in understanding and managing ecosystems. Among other issues, they are central to determining if an ecosystem is vulnerable to a particular stressor, evaluating the ability of alternate management actions to achieve environmental goals, designing an ecological restoration plan, evaluating the efficacy of risk-reduction measures such as remediation, assessing observed ecological responses to evaluate the necessity for adaptive management, and assessing when, and if, an ecological system has recovered from a past disturbance. Consequently, there is a broad-based need for appropriate, effective, and scientifically defensible tools for assessing ecosystem risks and recovery for managing human effects on the environment. One critical component of that need is the ability to assess risks to populations quantitatively.
We have argued elsewhere (Gentile and Harwell 2001; Gentile et al. 2001; Harwell 1997; Harwell et al. 1996; Harwell, Gentile, Cummins et al. 2010; Harwell and Gentile 2000) that assessing the condition and recovery of ecological systems in the presence of multiple natural and anthropogenic stressors requires a hierarchical approach, fundamentally based on the ecological risk assessment framework (USEPA 1992, 1998), focused on carefully selected ecological endpoints or valued ecosystem components (VECs), and cognizant of natural variability.
In Harwell and Gentile (2006), we applied these concepts to assessing the significance of any continuing effects on the Prince William Sound (PWS), Alaska, ecosystem more than 15 years after the Exxon Valdez oil spill (EVOS) (see also Harwell, Gentile, Cummins et al. 2010). In the ecological risk context, the issue is 2-fold, 1 pertaining to exposure and the other to effects: 1) Have the stressors caused by the oil spill decreased to the point where no further risk exists to the PWS ecosystem and its biota, and 2) Have VECs in PWS that were initially impacted now recovered?
The EVOS occurred on 24 March 1989, with over 250 000 barrels of Alaska North Slope crude oil released into northeastern PWS (Galt et al. 1991; Harrison 1991; NOAA 1992; Wolfe et al. 1994; Spies et al. 1996). Neff et al. (1995) reported that 782 km of the PWS shoreline (approximately 16%) was oiled to some degree after the spill. Four stressors resulted from EVOS: 1) volatile organic compounds (VOCs), which dissipated quickly but posed an early inhalation risk to some species, such as killer whales, 2) oiling, which caused loss of thermoregulation in the cold PWS waters and thus probably caused the majority of observed mortality to seabirds and sea otters, 3) polycyclic aromatic hydrocarbons (PAHs), which pose a longer-term toxicological risk to exposed organisms, and 4) the set of stressors caused by the cleanup activities of intertidal and shoreline habitat, including physical disturbance from high-pressure high-temperature water, extensive human presence, noise, water pollution, and wildlife disruption.
PWS is a very dynamic system, with extreme storms and high-energy wave and tidal regimes; consequently, residual Exxon Valdez oil (EVO) was largely eliminated naturally from shorelines in the initial months-to-few years after the spill, assisted by the massive cleanup operation during the summers of 1989–1991 (Teal 1991; Harrison 1991; Mearns 1996). By 2001 only approximately 0.1%–0.3% of the initial spill volume remained (Short et al. 2004), primarily in small, widely dispersed pockets of subsurface oil residues (SSOR) located in the mid-to-upper intertidal zone under a 15–25 cm-thick layer of clean sediments (Hayes and Michel 1999; Hayes et al. 2010), which in turn were located under a surface covering of stable armor composed of coarse gravel, cobble, and boulders (Hayes and Michel 1999; Michel et al. 2006; Taylor and Reimer 2008; Hayes et al. 2010). SSOR rarely occurred in unarmored, finer-grained sediments. Consequently, these subsurface deposits are released only very rarely and episodically into the coastal environment (Boehm et al. 2004; Neff et al. 2006) (otherwise, the residues would long since have left the system). For example, SSOR could be released by physical disturbance of the subsurface sediments during extreme storms, but after 2 decades, virtually all such storm-induced releases have already occurred. Similarly, SSOR could be released via bioturbation by sea otters' excavating pits when harvesting infauna; however, whereas this may have been a release mechanism in the early years after the oil spill, at present such SSOR excavations by sea otters occur only rarely (see the quantitative risk assessment in Harwell, Gentile, Johnson et al. ).
In addition to the physical release of SSOR, PAHs in the SSOR may gradually be released into the aqueous phase of the pore-water within the sediments and then transported to the surface water. As shown by Pope et al. (2011), however, such releases are at extremely low levels, limited by: 1) the very low solubility of remaining PAHs and the controlling equilibrium partitioning dynamics, 2) the very low permeability of the fine-grained sediments in which SSOR persist, and 3) the tidally driven, very large volumes of water that flow through the higher-permeability surface sediments, which greatly dilute any dissolved PAHs leached from subsurface deposits. As a result, these SSOR-derived PAHs in seawater are indistinguishable from background. However, in surface sediments, there are PAHs sorbed onto sediment particles that may become bioavailable if ingested by sediment-processors or filter-feeders (NRC 2003), including PAHs from pyrogenic sources, PAHs from other petrogenic sources, PAHs from biogenic sources, and, in some areas, low levels of PAHs that derived from EVO.
As a result, 3 of 4 classes of stressors resulting from EVOS, VOCs, oiling, and cleanup activities, were eliminated from the system many years ago. Presently the sole risk to PWS biota is the potential exposure to residual EVOS-derived PAHs, located either in the SSOR deposits or in the background levels of potentially bioavailable PAHs in EVOS-oiled areas.
Using the Gentile and Harwell (1998) criteria for assessing ecological significance (see also Harwell et al. 1994), Harwell and Gentile (2006) evaluated a suite of approximately 2 dozen VECs that characterize the PWS ecosystem, concluding that by that point in time recovery had essentially occurred for almost all of the VECs; similar conclusions were presented in Integral Consulting (2006), with a few exceptions. However, the Exxon Valdez Oil Spill Trustees (EVOSTC 2010) continued to report several endpoints as not recovered, including characterizing sea otters and Harlequin Ducks as “recovering” but not yet recovered species.
The sea otter (Enhydra lutris) is found in Alaskan coastal waters, feeding on benthic invertebrate epifauna and infauna, including clams, mussels, sea urchins, snails, and crabs (Bodkin and Ballachey 1997). Thus, its foraging habitat is limited in PWS to the relatively narrow zone along the shorelines that is sufficiently shallow for sea otters to reach the benthic communities (Bodkin and Ballachey 1997). At the time of EVOS, the PWS population of sea otters was rapidly expanding and increasing, as it was becoming reestablished throughout the northern Pacific after near-extinction by 1911, when the International Fur Seal Treaty ended harvesting (Doroff et al. 2003); for example, Estes (1990) reported the SE Alaska sea otter population increased at an annual rate of 17.6% from 1975 to 1987.
Their dense fur and unusually high metabolic rate make sea otters particularly vulnerable to oiling, both from decreased buoyancy and from hypothermia with loss of insulating capacity (Lipscomb et al. 1994), which is especially important because PWS is at the northern limit of the population range (Johnson and Garshelis 1995). Sea otters are also vulnerable to inhalation of toxic VOCs or ingestion of oil-contaminated food, exacerbated by increased rates of grooming after oiling (Johnson and Garshelis 1995). A total of 871 carcasses were found after EVOS, and 123 additional sea otters died in rehabilitation centers, totaling an observed mortality of approximately 1000 (Estes 1990; Loughlin et al. 1996; EVOSTC 2002). Garrott et al. (1993) estimated total sea otter mortality from EVOS in PWS at 2800. Johnson and Garshelis (1995) noted that by 1991 the sea otter population in the oiled areas equaled or surpassed prespill levels (based on 1984 estimates) at all their survey sites. Bodkin and Ballachey (1997) and Bodkin and Dean (2000) reported PWS sea otter abundance of 12 000–13 000 as surveyed in 1994, 1995, and 1999, exceeding the prespill estimates of 5000–10 000 (Burn 1994).
Despite an approximate 4% annual increase of the postspill sea otter population in PWS as a whole, an issue has been raised concerning the subpopulation of sea otters at Northern Knight Island (NKI). For instance, EVOSTC (2010) stated that although there had been a slow increase in the sea otter population at NKI since 2005, there had been a greater rate of overall increase in the PWS population as a whole, and therefore considered sea otters to be recovering but not yet recovered. Bodkin et al. (2002) and Dean et al. (2002) attributed the effects potentially to SSOR (no current effects are posited for the sea otters elsewhere in PWS, even though SSOR is not limited to NKI). Short et al. (2006) suggested that, when digging in sediments for food, NKI sea otters and seaducks may encounter SSOR in sufficient frequency and quantity to affect their health. In judging the plausibility of this suggestion, the issue, then, becomes one of quantitatively assessing the current risks from SSOR to NKI sea otters.
Similarly, the potential for long-term effects on the Harlequin Duck (Histrionicus histrionicus) population is the subject of continuing discussion. These are small seaducks that inhabit shallow marine intertidal zones off rocky shorelines during winter (Robertson and Goudie 1999). An estimated 500–1000 Harlequin Ducks (approximately 3%–7% of the wintering population) were killed as a result of direct exposure to EVO during and immediately after the spill (Esler et al. 2002; Rosenberg et al. 2005). Within weeks of the oil spill, concerns were expressed about potential persistent adverse toxic effects on Harlequin Ducks from ingesting oiled prey, especially mussels (Patten et al. 2000). Some studies conducted several years after EVOS have reported decreased abundance, attributing it to: 1) reduced overwintering survival of females and/or unsuccessful reproduction caused by EVOS (Esler et al. 2002; Patten et al. 2000), 2) degraded habitat quality not associated with EVOS (Day et al. 1995, 1997; Irons et al. 2000; Wiens et al. 2004), or 3) exposure to hydrocarbons and, by inference, continuing toxic effects from EVOS, as indicated by elevated cytochrome P4501A [CYP1A] activity (Trust et al. 2000; Esler et al. 2002; Esler 2008; Esler and Iverson 2010). Based on a weight-of-evidence evaluation from several studies, Wiens et al. (2010) concluded no indication of continuing population-level impacts, in terms of either abundance or demographics, on Harlequin Ducks from EVOS. On the other hand, EVOSTC (2010) concluded that there continues to be a risk of exposure to hydrocarbons from EVOS 2 decades after the oil spill, based in part on the CYP1A data (see also Esler 2008). Consequently, EVOSTC (2010) classifies Harlequin Ducks as “recovering but not recovered,” based on the recovery objective that “…biochemical indicators of hydrocarbon exposure in harlequins in oiled areas of PWS are similar to those in harlequins in unoiled areas.” We have argued elsewhere (Harwell and Gentile 2006; Harwell et al. 2012) that biochemical markers are inappropriate to use as recovery objectives, because biomarkers only can indicate exposure, not effects, and recovery by definition is an effects issue. Nevertheless, the relevant issue here becomes one of quantitatively assessing the current risks from EVOS to PWS seaducks to answer the question, is it possible or not for the residual PAHs from EVOS to cause effects on seaducks?
To assess the possibility that residual PAHs from EVOS could continue to cause adverse effects on those 2 species, Harwell, Gentile, Johnson et al. (2010) and Harwell et al. (2012) conducted quantitative ecological risk assessments (ERA) on PWS sea otters and seaducks, respectively. In each case, those authors developed a conceptual model of all plausible pathways of PAH exposures to the animals; converted the conceptual models into quantitative, stochastic, individual-based models (IBMs), parameterized with empirical data on those species as well as data on the SSOR and other EVOS-derived PAHs in PWS; developed chronic toxicity reference values (TRVs) for PAHs appropriate for sea otters and seaducks, based on the best-available applicable data in the literature; simulated millions of sea otters and seaducks to derive assimilated doses of PAHs; and assessed the frequency distributions of exposures and resulting effects to reach conclusions on the magnitude of the potential risks to individual sea otters or seaducks.
In both studies, the authors assessed the chronic doses to the 99.9% quantile individuals (i.e., the 1-in-1000th most-exposed individual sea otters or seaducks). This approach, coupled with several other conservative attributes built into the models, provides a very conservative estimate of risks. The results indicated that the assimilated doses for sea otters were approximately 30–125 times lower than the no-observed-adverse-effects-level (NOAEL) TRV threshold, and approximately 75–310 times lower than the lowest-observed-adverse-effects-level (LOAEL) TRV threshold. For the Harlequin Ducks, the assimilated doses were approximately 400–4000 times lower than the NOAEL and LOAEL TRVs, respectively. Based on these results, the authors concluded that there was essentially no remaining individual-level risk to sea otters or seaducks from EVOS (Harwell, Gentile, Johnson et al. 2010; Harwell et al. 2012).
In the present study, we extend the use of the individual-based models to examine the potential for population-level effects. Under present PAH exposures, no individual-level effects are plausible and thus no population-level effects could occur. However, in a hypothetical environment in which the PAH concentrations are heuristically increased to attain TRV levels, the IBM-class of models provides a useful tool to assess population-level risks, a tool that would equally be useful for situations in which actual exposures are close to toxicity threshold values. The stochastic IBMs allow the generation of frequency distributions that represent the range of possible effects on individuals by capturing the underlying variability in exposures. These frequency distributions can be used to project the environmental exposures that would be necessary to cause detectable effects in populations. We believe the approach used here would be generally applicable to using stochastic individual-based models to quantitatively assess population-level risks.