To identify human activities, pressures, and ecosystem services relevant to risk assessment, we proposed the Driver–Activities–Pressures–State–Impact (Welfare)–Response (Measure) (DAPSI(W)R(M)) framework alongside the UN Common International Classification of Ecosystem Services (CICES) framework. The DAPSI(W)R(M) framework provides a basis for selecting pressures and activities, and for establishing causal links between these activities, pressures, and the ecosystem studied. In contrast, the CICES framework helps classify the ecosystem services impacted by these activities and pressures.

The DAPSI(W)R(M) framework is a recent and widely recommended conceptual model for marine EBM (Elliott et al., 2017). It refines the Driver–Pressure–State–Impact–Response (DPSIR) approach by addressing key limitations, enabling the identification of more holistic management strategies that can address the linkages between various environmental problems and their associated drivers and pressures (Cooper et al., 2013; Atkins et al., 2011). Within the DAPSI(W)R(M) framework, drivers refer to societal drivers that are related to basic human needs, such as food, energy, and security. These drivers lead to activities (A)—human intervention—that generate pressures, which are the mechanisms of change. These pressures, in turn, cause alterations in the state (S) of the natural system, resulting in certain impacts (I) (on welfare) on the social system that prompt various responses (measures), such as prevention, mitigation, compensation, or other initiatives. Examples of pressures in a marine context include pollution, climate change, and invasive species. The I (on welfare) are measured here as the risk associated with changes in the state of the environment affecting both biotic and abiotic ecosystem services. For the pragmatic application of DAPSI(W)R(M) in this study, we made some simplifications. First, we defined the overall health of the marine ecosystem as the sole indicator of the state. Second, we focused on identifying human activities and natural pressures that can be clearly observed and are indisputable, rather than considering all societal needs of drivers, which may be disputable. Third, we prioritized negative impacts in this study, as these are often highlighted and receive significant attention due to their urgency and the need for intervention. In contrast, positive impacts may be uncertain or difficult to quantify, which presents challenges in establishing causation and thus falls outside the scope of this study.

Ecosystem services were first broadly introduced by the Millennium Ecosystem Assessment (MEA) and defined as “the benefits that people obtain from ecosystems”, both natural and managed (MEA, 2005). These services are categorized into provisioning, regulating, cultural, and supporting services. The first three categories have a direct impact on human well-being, while the latter has an indirect impact by supporting provisioning, regulating, and cultural services. Since its release, the MEA has served as a conceptual framework identifying, classifying, and quantifying ecosystem services (ESs) (Armstrong et al., 2012; Costanza et al., 2017; Fisher et al., 2009). Other more recent frameworks, including The Economics of Ecosystems and Biodiversity (TEEB) (TEEB, 2010), the CICES (CICES, 2013), and the Intergovernmental Platform on Biodiversity and Ecosystem Services (IPBES) (IPBES, 2017), have been developed to differentiate, structure, and provide the basis for evaluating ecosystem services (Thurber et al., 2014). To some degree, these frameworks critique the MEA framework, particularly with regard to the inclusion of both direct and indirect services, which may create the potential for double counting values.

We applied the CICES framework (Haines-Young, 2023) because it represents an important step forward compared to earlier classifications provided by MEA (MEA, 2005) and TEEB (TEEB, 2010). First, in CICES, ESs are defined as the contributions ecosystems make to human well-being, distinct from the goods and benefits people derive from them. This framework emphasizes identifying both the purposes and uses people have for the different kinds of ESs. Second, supporting services are excluded from the classification to avoid the issue of “double counting” when linking ecosystem and economic accounts. Third, CICES distinguishes between ESs that depend on living systems (i.e., biophysical ecosystem outputs) and those that involve nonliving (abiotic) parts of ecosystems, which also contribute to human well-being (geophysical ecosystem outputs). The framework also facilitates a clearer identification of the different service categories and their coverage. For this study, we defined and classified all the biotic and abiotic services examined, focusing on the class level of the CICES framework.

Risk matrices are widely adopted structures for risk assessment in marine ecosystem-based management (Cox, 2008). They are commonly used to provide a ranking framework for visualizing and prioritizing risks, guiding resource allocations in a convenient and efficient manner. Different concepts of risk have been discussed in recent years to address the complex interactions and uncertainties associated with climate change and human activities (Oppenheimer et al., 2015; Field et al., 2012; Hobday et al., 2011). In the most common definition, risk is the product of likelihood and effect. Specifically, risk is typically seen as the product of (1) the likelihood that an event will occur, and (2) the effect that event may have. However, recent literature emphasizes the potential pitfalls of the risk matrix, such as risk rating, limitations in resolution, and the possibility of errors (Cox, 2008). Additionally, the risk matrix establishment process often fails to account for the risk attitudes of decision-makers (Ruan et al., 2015; Kaya et al., 2019). Therefore, risk should be viewed as a multivariate construct rather than a scalar. Furthermore, to evaluate risk in a more comprehensive manner, it is recommended that it be viewed in a way that integrates risk attitudes based on utility theory (Ruan et al., 2015).

In this study, we have nonetheless chosen to use the risk matrix method for several reasons. First, despite its limitations, this approach is still widely applied in many fields (Cox, 2008; Duijm, 2015). Second, assessing risk based on utility theory may also be problematic due to what is commonly known as the Allais paradox (Allais, 1953). For instance, people may not always make rational decisions, i.e., based on their utility, but may instead be influenced by cognitive biases that affect their perception of risk (Zhang et al., 2022). This is due to social and cultural relationships, meaning that the opinions of others can influence one’s choices (Nick et al., 2019).

We designed the risk matrix based on the principles of weak consistency, betweenness, and consistent coloring, as recommended in more recent literature (e.g., Cox, 2008; Li et al., 2018). Solutions for addressing limitations such as poor resolution, range compression, and ranking reversal errors, which aim to minimize their impacts, are also considered ( Figure 1 ). In addition, we ensure that all risk values associated with higher-rated cells are quantitatively greater than those in lower-rated cells. To avoid the issue of scalar risks, we present risk as a graphical combination of likelihood and effect, offering a more comprehensive understanding of the risks involved.

In recent years, there has been increased emphasis on uncertainties concerning risk assessment and management (Brown and Cox, 2011; Aven and Zio, 2011). Criticisms have been directed at the traditional probabilistic approach to risk assessment and management due to its limited scope in addressing risk and handling uncertainties. As proposed by Aven (2017), the general framework for risk description should encompass three components: effect (E), likelihood (L), and background knowledge (K), upon which both E and L are based. This approach to describe risk allows for all types of uncertainty representations and could serve as a basis for a unified perspective on uncertainties within the context of risk assessment.

Following this recommendation, we conduct a certainty assessment in addition to the risk assessment for the study. This allows us to examine whether the lack of effect and likelihood arises from insufficient knowledge (uncertainty) or from a well-understood probabilistic process (risk). Certainty is assessed in the same manner as the risk assessment, i.e., based on experts’ judgment using a Likert scale ranging from very uncertain (1) to very certain (5) for all the drivers and ecosystem services.

Despite the growing recognition of the importance of describing uncertainty in risk assessment, there remains a scarcity of practical proposals for assessing uncertainty in relation to risk. One alternative qualitative approach is visualizing uncertainty in risk diagrams, which includes concepts such as the family of risk curves (e.g., Kaplan and Garrick, 1981), uncertainty boxes in probability-consequence diagrams (e.g., Duijm, 2015), and bubble diagrams (e.g., Abrahamsen and Aven, 2011). Recent recommendations favor the use the three-dimensional PCD-PISEA visualization (probability-consequence diagrams with prediction intervals and strength-of-evidence assessments). This method incorporates three dimensions: (i) the assigned likelihood for event occurrence (L), (ii) a prediction for the effect given the occurrence of events (E), and (iii) a measure of the strength of evidence upon which the likelihood and effect assignments are based (K). Here, we propose a three-dimensional visualization using a spider web diagram for judgment. Furthermore, we investigate whether certainty levels for ranking risk between services and between drivers differ significantly from each other. To do this, we perform a pairwise Wilcoxon signed-rank test with two types of correction for multiple testing: Bonferroni correction and False Discovery Rate (FDR). While Bonferroni’s correction is conservative, FDR helps avoid false-positive outcomes. This approach is appropriate because the assumption of a normal distribution is not met for the data.

The Barents Sea region is a high-latitude shelf ecosystem situated north of Norway and Russia ( Figure 2 ). It covers an area of approximately 1.6 million km² and has an average depth of around 230 m. The region hosts a wealth of resources, including oil and gas, minerals, and some of the largest fish stocks in the world. With 5.1 million people living around the Barents Sea and extensive international trade and tourism, a broad range of human activities take place in this region.

The management plan for the Barents Sea was first presented as a white paper titled Integrated Management of the Marine Environment of the Barents Sea and the Sea Areas off the Lofoten Islands (Report No. 8 (2005–2006) to the Storting) (The Royal Norwegian Ministry of the Environment, 2006). This plan has since been revised, with the most recent update in 2011, in the white paper First update of the Integrated Management Plan for the Marine Environment of the Barents Sea–Lofoten Area (Meld. St. 10 (2010–2011) Report to the Storting) (Norwegian Ministry of the Government (2012)). The plan consolidates existing and new knowledge about the main activities, pressures, ecosystems, and services, as well as trends in regional risks. The main activities and pressures in the Barents Sea were also summarized in The Professional Forum for Norwegian Marine Areas (2023).

Accordingly, primary activities in the Barents Sea include fishing, tourism, oil and gas, and shipping. The region is regarded as one of the most productive fishing areas in the world, home to major commercial species such as cod, haddock, and capelin. It is also rich in hydrocarbons, leading to increased oil and gas exploration and extraction activities. This includes drilling operations, seismic surveys, and the establishment of offshore platforms, all of which have raised environmental concerns due to increased pollution, noise, and the risk of accidental spills. As climate change reduces ice coverage, the Barents Sea is becoming more accessible for both petroleum activities and shipping routes. Increased shipping traffic, including cargo vessels and cruise ships, poses potential risks to marine ecosystems (e.g., pollution, vessel noise, accidental spills) and safety. Invasive species, such as the king crab and snow crab, have also been recognized in the region (Husa et al., 2022).

The Barents Sea provides numerous ecosystem services, including the supply of seafood and oil and gas resources, which are vital for local and regional economies. The region plays an important role in climate regulation, carbon sequestration, and water purification (Rogge et al., 2023; Smedsrud et al., 2013). It is also significant for cultural identity, both locally and nationally, and supports educational and research opportunities that enhance public awareness of marine conservation.

However, the region is also impacted by global stressors, such as climate change and long-range pollution. Climate change is evident in the Barents Sea through rising sea temperatures, which alter the distribution, growth rates, and reproductive cycles of marine species. Shrinking ice cover affects habitat availability for ice-dependent species, leading to shifts in food webs and ecological dynamics (Cooley et al., 2022; Mikkelsen et al., 2023). Pollutants, including heavy metals, microplastics, and organic contaminants, are increasingly entering the Barents Sea due to local activities and long-range atmospheric transport. These pollutants threaten marine life and can accumulate in the food chain, posing health risks for humans who depend on marine resources (Frantzen et al., 2022).

The survey was conducted by collecting experts’ opinions to identify relevant human activities, pressures, and ecosystem services at risk in the Barents Sea, followed by an evaluation of the risks posed by these different pressures and activities on the ecosystem services. The survey was developed specifically for this case, though it was based on a previous Atlantic Ocean deep-sea ES risk assessment (Armstrong et al., 2019).

As already mentioned, the data are limited, and there is significant variance in the number of responses for each activity, pressure, and service, making it challenging to compare their likelihood, effect, and consequently, risk levels. We have therefore attempted to assess the lack of response, i.e., the “holes” (or the zero values) in the dataset, to illustrate the differences in knowledge among the respondents across services, activities, and pressures. This is illustrated in Figure 3 , which shows a greater number of scored values for the services fish/shellfish and biodiversity, while educational services have more “nonapplicable” values. We also observe that invasives and shipping are marked as “nonapplicable” for most ESs, whereas pollution and temperature change are on the opposite end of the scale.

In the following, we present the results from two forms of risk associated with pressures and anthropogenic activities identified in the Barents Sea: (1) the risk associated with each activity and each pressure ( Figure 4 ) and (2) the risk resulting from each ES ( Figure 5 ). In Figure 4 , the general impression is that all pressures and activities generate risks. Although the levels of risk varied among the pressures and activities depending on the ES being assessed, all risks were medium or low. However, some distinct results emerge. For instance, shipping has the lowest variance in risks, whereas temperature change shows the greatest variance overall. Although temperature change is recognized as a medium-risk pressure, changes in temperature poses higher risk levels to a greater number of ESs than other pressures. Meanwhile, pollution poses a medium risk to most ESs and, in some cases, a low risk. The same situation applies to fisheries and oil and gas activities. Shipping is the only activity that places all ESs at a low-risk level.

Figure 5 focuses on the services, illustrating that all are at risk of impact from a broad spectrum of activities and pressures; however, the risk levels vary among them. Notably, fish/shellfish and biodiversity services are ranked as the highest-risk services under most activities and pressures. To a lesser degree, biological control, climate regulation, and existence/bequest are also highly ranked, while educational services appear to be much less at risk. Interestingly, perceived risk levels for fish/shellfish from, e.g., pollution are higher than those from fisheries, suggesting that the respondents may have high confidence in fisheries management. However, the difference in risk to fish/shellfish between pollution and fisheries is not statistically significant.

As seen in Figure 6 , we observe that all services, pressures, and human activities are perceived at medium certainty on average, with somewhat less certainty associated with ESs compared to the pressures and activities. The mean values differ slightly from the median values in some services/pressures and activities. For instance, all ESs, pressures, and activities have the same certainty mean of 2, with a standard deviation of ± 0.9 for ecosystem services, ± 1.3 for activities, and ± 1.1 for pressures. Meanwhile, the median certainty for ES is 2, while it is 3 for pressures and activities. This is as expected, given that the data are not normally distributed.

Among the identified pressures, temperature change is ranked as having the highest certainty levels, while invasives exhibit the lowest. Experts are also fairly certain about pollution, although this certainty is somewhat lower than that associated with temperature change. Regarding activities, fisheries are perceived to have the highest certainty levels, whereas shipping is associated with significantly lower certainty.

With respect to ecosystem services, certainty varies between the services, and its variance is generally high. The highest certainty levels are assigned to fish/shellfish, followed by climate regulation and biodiversity. The certainty levels for oil and gas/energy, raw materials, waste absorption, and carbon sequestration/absorption are also significantly different from most of the services identified in the Barents Sea.

The statistical results (p-values) of the Wilcoxon test to determine significant differences in the levels of certainty between the pressures, human activities, and services are presented in Tables 4 , 5 . Two methods to avoid multiple comparison problems, Bonferroni correction and FDR, are applied to detect significant differences in certainty levels (when scoring risk) between services/pressures and activities. The certainty levels of temperature change and fisheries are similar but significantly different from the certainty levels of oil and gas, shipping, and invasives. While the certainty level of pollution is similar to those of temperature change and fisheries, it is not significantly different from the level of oil and gas. For services, those significant p-values support that the certainty levels of WD, Edu, EB, CSA, and BC are similar to those of FS, OGE, CR, and Bio, as seen in Figure 6 . The level and quantile ranges of WA appear similar to those of RM in plots, but the p-value indicates a significant difference. The results may be due to a number of outliers.

Figures 7 , 8 present risks in relation to certainty for services, pressures, and activities using radar charts. While the certainty levels are assumed to be the same for all pressures and activities in each ES ( Figure 7 ) and for all services in each pressure or activity ( Figure 8 ), the likelihood and effect vary accordingly. As seen in Figure 7 , the likelihood, effect, and certainty scores for fish/shellfish and biodiversity are the greatest, suggesting these are the most threatened services with the highest certainty. Climate regulation is also perceived to be among the greatest at-risk services, at least in relation to “temperature change”. In contrast, education is perceived as the least threatened service, albeit with the lowest certainty. In Figure 8 , temperature change and pollution are recognized as the most threatening pressures, with higher certainty levels for the services of Bio, CR, FS, and OGE, and lower certainty levels for BC, CSA, EB, WD, WA, RM, and Edu. Shipping is considered the least threatening activity with the lowest certainty. This result may reflect how well temperature change and pollution have been studied in the literature. A caveat is, of course, that most of the respondents are natural scientists who work on topics related to fisheries, climate regulation, and biodiversity.

Addressing all identified risks is challenging due to resource limitations and inherent trade-offs between ESs. This raise the question: where should greater management efforts be focused. Should priority be given to ecosystem services with the highest perceived risks? Or could it be that when we have better knowledge, we feel more confidents about assessing the risks? Our assessment reveals that fisheries is not identified as the highest-risk activity in this study, despite having the greatest certainty among activities. Interestingly, fisheries is often ranked as a top-risk activity in other studies (e.g., Hansen et al., 2022; Pedreschi et al., 2023). This demonstrates strong confidence in the scientific foundation of the fisheries management system in the Barents Sea. Extensive efforts in science and management—more than for any other marine sector (e.g., annual fish stock monitoring surveys, annual stock assessments, and a well-developed system for management advice and harvest control rules)—have contributed to a successful transition to sustainable harvesting of the dominant fish stocks in the Barents Sea over the recent decades (Gullestad et al., 2017).

Meanwhile, fish/shellfish production and biodiversity services face the highest risks from temperature change and pollution pressures, which are associated with high certainty levels. This suggests that management strategies should prioritize these areas to mitigate risks effectively. These strategies could include investing in an advanced monitoring systems to detect early signs of temperature anomalies and pollution levels, enabling timely management; strengthening regulations to control pollutants entering marine ecosystems; and encouraging sustainable fishing practices that minimize bycatch and habitat destruction. Additionally, raising public awareness about the impacts of temperature change and pollution on marine ecosystems can promote community involvement in conservation efforts.

The study found that fisheries-associated risk was scored slightly higher for biodiversity than for fish/shellfish, aligning with Siwertsson et al. (2023). This demonstrates that fisheries affect the broader ecosystem, causing significant bycatch by capturing nontarget species and habitat destruction, rather than just impacting the targeted species. Conservation efforts may need to strengthen the focus on protecting marine biodiversity in addition to managing the target fish and shellfish stocks. This includes implementing additional measures to reduce bycatch, protect critical habitats, and enforce sustainable fishing practices. Management strategies should focus on ecosystem-based approaches rather than single-species management. This could involve setting biodiversity conservation targets and enhancing monitoring and enforcement to ensure compliance with sustainable fishing regulations. Policies and regulations may need to be revised to address the broader ecological impacts of fisheries. This could include stricter controls on fishing gear and methods, catch limits that consider ecosystem health, and more robust environmental impact assessments before allowing fishing activities. Recent efforts along these lines include the establishment of areas closed to fisheries in the northern Barents Sea, which are becoming more accessible to fisheries due to climate warming and receding sea ice (Jørgensen et al., 2020).

The results also point to challenges for management. It is, therefore, essential to define the managed system we wish to achieve, balance trade-offs between ESs, determine which services, pressures, and activities should be prioritized, and identify the necessary risk mitigation measures in practice. The prioritizations depend on both the management level and spatial scale. Local managers may focus on high-risk, high-certainty issues, where immediate actions can be implemented effectively. In contrast, regional authorities are tasked with balancing these with longer-term, uncertain risks, requiring resource allocation between direct interventions and research. At the international level, cooperation becomes crucial for addressing high-risk, low-certainty issues that cross boundaries, necessitating collaborative frameworks and shared resources to manage complex, transboundary challenges. Such multilevel management systems ensure a comprehensive approach to risk management.

In the context of this study, fisheries represent a significant risk activity with a high degree of certainty. As such, national managers are advised to prioritize this area, as they have the authority to establish regulations, such as closing areas to fisheries in general or restricting the use of specific gears, such as bottom trawling, which impacts benthic habitats within areas of Norwegian jurisdictions (e.g., Jørgensen et al., 2020). Additionally, pollution and temperature change were identified as primary risk pressures for the Barents Sea region. Although these pressures were measured as having lower levels of risk and certainty, their transboundary nature suggests that addressing them requires longer-term efforts, including national or international cooperation for effective risk mitigation.

The study identifies significant knowledge gaps. The medium and low certainty perceived in all activities, pressures, and ESs indicates limited knowledge of risks in the Barents Sea. This finding is consistent with Siwertsson et al. (2023). Further research is needed to enhance understanding of ESs in relation to activities and pressures, as certainty regarding ESs is generally lower. This involves gathering more comprehensive data and developing advanced models to predict the risks of various pressures and activities on ESs more accurately. Specifically, for carbon sequestration and oil and gas/energy services, which are high risk but low certainty, more in-depth studies are necessary to pave the way for risk mitigation strategies. Since fisheries are seen as impacting biodiversity more than fish/shellfish, ongoing research and monitoring are needed to better understand the impacts of fisheries on biodiversity. This includes studying the cumulative and indirect effects of fishing on nontarget species and habitats and developing indicators for ecosystem health.

It should be emphasized that the findings are based on the qualified assessments of a limited number of respondents, predominantly from a natural science background. Therefore, the results should be interpreted with caution. Variations in disciplinary backgrounds may lead to different risk perceptions, and imbalanced representation may skew overall results, especially if certain perspectives are under- or over-represented. It would be desirable to analyze the influence of different disciplinary backgrounds on our results, as this would provide valuable insights into how these disciplines influence responses to identification and risk perceptions. Additionally, examining whether project member involvement affects scoring and how interactions during the workshops influence identification and scoring would further enrich the analysis. Unfortunately, the limited number of respondents does not permit such analyses. In this study, a more balanced representation of expertise, particularly from social scientists, is often more desirable, as ecosystem services bridge the natural and social sciences. Ecosystem services are human-centric in focus, with interdisciplinary perspectives. They incorporate knowledge from economics, sociology, anthropology, and political science to address questions about equity, governance, and resource management. Social scientists bring important knowledge relevant to addressing conflicts, trade-offs, and synergies among stakeholders with different interests, as well as insights into the role of governance structures, economic incentives, and policy frameworks in managing ecosystem services. This is especially true for regulating and cultural services, which often have less direct market value but are of significant societal importance.

In this study, the Barents Sea is treated as a homogeneous area, which may not capture localized variations in ecosystem processes or site-specific management decisions. For instance, the risk to marine ecosystems and likely to ecosystem services increases from the open ocean toward the coastal areas due to increasing human activities and impacts (Hansen et al., 2022; Aarflot et al., 2024). While this simplification may affect the accuracy of localized risk estimates, it enables a comprehensive regional overview. Future studies should incorporate spatially explicit data to account for these variations and align the analysis with localized management needs. Additionally, the study focused on individual risks only, but it is important to consider the indirect and cumulative risks of multiple activities and pressures on ESs and their interactions. For instance, the indirect risks associated with temperature change, such as marine heatwaves and policy changes, can substantially impact the ecosystem. These cumulative effects can exacerbate the overall risk to ESs, leading to more complex and potentially severe consequences than those posed by individual activities or pressures alone.