Oil spill risk analysis in the Arctic

In our work, we apply the general concept of risk, which combines the probability of an oil accident with the estimated environmental impacts of the oil spill. The approach includes three steps illustrated in the below figure. First, we estimate the population at risk, i.e. the proportion of the population of a given species that occurs in the oiled area after the accident. This is done by combining probabilistic maps describing the relative densities of species with the estimated extent of oil slick. Second, the proportion of the population that will actually die due to the spilled oil is estimated by taking into account the species-specific exposure potentials and sensitivities to oil. In the third step, the estimated impacts are combined with the spatio-temporally varying oil accident probability. All steps are modelled by using probabilistic methods, which enables taking uncertainty into account explicitly. Our risk assessment model is summarized in the below figure.

Figure. The conceptual framework of Artic oil spill risk analysis developed in our group. On the left, a directed acyclic graph (DAG) showing the factors in the oil spill risk assessment and their relationships. On the right, a description of the factors: (A) Factors that can be controlled by management decisions; (B) seasonally and spatially varying factors.  (C) Calculation of the expected proportion of a population that is at risk. (D) Spatially constant but seasonally varying species-specific exposure potential and sensitivity and seasonally varying route-specific oil spill impact. The figure is reproduced from Environ. Sci. Technol. 2020, 54, 4, 2112-2121. https://doi.org/10.1021/acs.est.9b07086

The abundance of species, the spreading capacity of oil spill and the accident probability depend on the environmental conditions in the accident location and around it. We derived environmental covariate data from publicly available data repositories, which are maintained by international and foreign science institutes, such as NASA. The covariate data covered sea ice cover and depth, which are available as spatially continuous maps. In addition to them, we were interested in the salinity content of the sea surface, since it indicates the fresh water inflow from the Eurasian continent to the Arctic Ocean and positively correlates with primary production as carrying high loads of nutrients. Oil spreading capacity and accident probability depend solely on the ice cover but the distributions of species depend on the depth and fresh water distribution as well. Thus, we needed to create spatially continuous maps of sea surface salinity. We derived point observations of sea surface hydrography, namely temperature and salinity, from publicly available data repositories. The point observations were input for statistical spatio-temporal models, which we used for predicting the hydrographic conditions across the study area. The prediction maps of sea surface hydrography were used to examine whether the Arctic continental shelf had undergone a shift in the average hydrographic conditions during the time period 1980-2000. Furthermore, they were used to link species observations to the sea surface salinity content.

Figure. Distribution of temperature and salinity data in the Kara Sea. These data were used for analysing spatiotemporal changes in temperature and salinity across Kara Sea (Mäkinen and Vanhatalo, 2016) and later on to construct environmental covariate layers for species distribution modeling.

We studied spatial distributions of polar bears, walruses and ringed seals to track changes in the distributions during the recent decades and to provide spatially accurate information of species abundance for oil spill scenarios. For both tasks, we needed spatially continuous maps of species areal abundances. Such maps are scarce in the Arctic areas, which suffer from low survey efforts for mapping species distributions and populations. Moreover, the Eurasian side of the Arctic suffers from a lack of any kind of observational data of the studied species. Thus, we looked from published books and reports for such observational data, which have not been digitized and put publicly available. The search led us to reports of Russian expeditions, which provided data for creating species distribution models. For the models, we linked the species observations to environmental conditions based on the locations and temporal stamps of the observations. The models were used to infer the relationship between abundance of each species and the environmental conditions. Moreover, the models needed to account for unmeasured ecological processes and varying survey effort between data sources. The inferred relationships were used to predict species abundance across the study area on a monthly resolution in the time period covered by species observations (1996-2013) to track changes in distributions. Moreover, they were used to predict species abundance along the marine traffic lines to provide input data for oil spill scenarios.

Figure. Example of species data used in our studies. Each line denotes a survey transect through which an  expedition had travelled. The dot's denote the locations of species sightings. These information were compiled and used to train a species distribution model (Mäkinen and Vanhatalo, 2018) with which we then predicted the spatiotemporal changes in species abundance.

Figure. The avarege seasonal distribution of polar bears, walrus and ringed seals in the Kara Sea and the relative change in spring density from 1990's to 2000's. Figure is reproduced from Mäkinen and Vanhatalo (2018): https://doi.org/10.1111/ddi.12776.

To quantitatively assess these risks that oil poses to Arctic animals we need also knowledge about their exposure potential and sensitivity to spilled oil. However, in the Arctic these data are typically scarce or lacking altogether. To compensate for this limited data availability we conducted probabilistic expert elicitation to estimate these parameters for seals, anatids, and seabirds. Our results suggest that the exposure potential and sensitivity to oil vary between functional groups, seasons, and oil types. Overall, the impacts are least for seals and greatest for anatids. Moreover, offspring seem to be more sensitive than adults. The elicitation tool and results were presented by Nevalainen et al. (2018, 2019)

We have applied the method to the Kara Sea, the westernmost part of the Northern Sea Route, where we have estimated the impacts of a potential oil spill on three marine mammal species (polar bear, ringed seal and walrus). The assessment is conducted for five shipping routes, four oils types and three seasons.

Figure. The case study area (Kara Sea) and the shipping routes examined by Helle et al. (2020). The figure is reproduced from Environ. Sci. Technol. 2020, 54, 4, 2112-2121. https://doi.org/10.1021/acs.est.9b07086

To estimate the population at risk (step 1) in the Kara Sea, we use probabilistic maps that describe the relative densities of three arctic marine mammal species in relation to the relevant environmental covariates, spatial location and time. The maps are based on Bayesian hierarchical species distribution models and data gathered from scientific publications and open source data repositories. Further, the size of the oil slick (step 1) is assumed to be dependent on the properties of oil and ice conditions. The species-specific exposure potentials and sensitivities (step 2) are based on expert elicitation, as field and laboratory data on the topic are scarce. In the near future, the analysis will be supplemented with the accident probabilities along the shipping routes (step 3).

Our analysis shows that the impacts of potential oil spills differ considerably between species, shipping routes, seasons and oil types. Although medium and heavy density oils may be more detrimental to the biota than light oils, the latter have potential to pollute larger areas than the former ones. The uncertainties associated of the estimates are typically large and stem from high environmental variability and lack of detailed data. Hence, there is a need for improved understanding on the species distributions and abundances in the Arctic, on the environmental factors affecting the fate of oil in ice conditions, and on the species-specific behavioral and physiological factors that determine the final impact of the oil exposure.

Figure. Average proportion of population at risk (AvgPPRs) and average oil spill impact (avgOSIs) for polar bears, walrus and seals along routes R1–R5 with medium (rows 1 and 3) and extra heavy (rows 2 and 4) oil in spring (green), summer (orange), and autumn (purple). Circles denote medians; thick lines are the 25% to 75% quantile, and thin lines the 5% to 95% quantile. The figure is reproduced from Environ. Sci. Technol. 2020, 54, 4, 2112-2121. https://doi.org/10.1021/acs.est.9b07086

More information on our research related to oil spill risk assessment in the Arctic:

General framework and combined analysis:

• Helle, I., Mäkinen, J., Nevalainen, M., Afenyo, M. and Vanhatalo, J. 2020: Impacts of Oil Spills on Arctic Marine Ecosystems: A Quantitative and Probabilistic Risk Assessment Perspective. Environ. Sci. Technol. 54(4): 2112–2121. DOI: 10.1021/acs.est.9b07086
• Nevalainen, M., Helle, I. and Vanhatalo, J. 2017. Preparing for the unprecedented - Towards quantitative oil risk assessment in the Arctic marine areas. Mar. Pollut. Bull. 114(1): 90– 101. DOI: 10.1016/j.marpolbul.2016.08.064

Environmental factors and species distribution modeling:

• Mäkinen, J. and Vanhatalo, J. 2018: Hierarchical Bayesian model reveals the distributional shifts of Arctic marine mammals. Divers. Distrib. 24 (10): 1381– 1394. DOI: 10.1111/ddi.12776
• Mäkinen, J., Vanhatalo, J. 2016: Hydrographic responses to regional covariates across the Kara Sea. J. Geophys. Res. - Oceans 2016 121(12): 8872– 8887. DOI: 10.1002/2016JC011981

Vulnerability and sensitivity of the Arctic biota to oil:

• Nevalainen, M., Vanhatalo, J. and Helle, I. 2019: Index-based approach for estimating vulnerability of Arctic biota to oil spills. Ecosphere 10(6): e02766. DOI: 10.1002/ecs2.2766
• Nevalainen, M., Helle, I. and Vanhatalo, J. 2018: Estimating the acute impacts of Arctic marine oil spills using expert elicitation. Mar. Pollut. Bull. 131: 782–792. DOI: 10.1016/j.marpolbul.2018.04.076

Ice loads on ship hull and probability of shipping accidents

• Jarno Vanhatalo, Juri Huuhtanen, Martin Bergström, Inari Helle, Jussi Mäkinen and Pentti Kujala (2021). Probability of a ship becoming beset in ice along the Northern Sea Route - a Bayesian analysis of real-life data. Cold Regions Science and Technology, 184:103238. DOI: https://doi.org/10.1016/j.coldregions.2021.103238
• Kotilainen, M., Vanhatalo, J. Suominen, M. and Kujala, P. (2018). Predicting Local Ice Loads on Ship Bow as a Function of Ice and Operational Conditions in the Southern Sea.. Ship Technology Research - Schiffstechnik 65(2):87-101. DOI: 10.1080/09377255.2018.1454390
• Kotilainen, M., Vanhatalo, J. Suominen, M. and Kujala, P. (2017). Predicting ice-induced load amplitudes on ship bow conditional on ice thickness and ship speed in the Baltic Sea.. Cold Regions Science and Technology, 135: 116-126. DOI: 10.1016/j.coldregions.2016.12.006

PhD theses produced during the projects

• Nevalainen M. (2019). Preparing for the unprecedented –Moving towards quantitative understanding of oil spill impacts on Arctic marine biota. E-thesis. News
• Mäkinen J. (2020). Geographies of uncertainty - methodologies for detecting environmental and ecological changes. E-thesis.