A Modified PESTEL- and FCM-Driven Decision Support System to Mitigate the Extinction of Marine Species in the Mediterranean Sea

Abstract

The Mediterranean Sea, a biodiversity hotspot with over 8500 marine species, faces escalating threats from climate change, pollution, overfishing, and habitat degradation. This study introduces a novel Decision Support System (DSS) integrating a modified PESTEL framework (BESTEL: Biological, Economic, Social, Technological, Environmental, Legal) with Fuzzy Cognitive Mapping (FCM) to assess and mitigate risks to marine species. Leveraging expert knowledge from 34 specialists, we identified 30 key factors influencing extinction risk, analyzed through Principal Component Analysis (PCA) to reduce dimensionality. The resulting FCM model simulated four policy scenarios, evaluating the impacts of climate change and dam proliferation on biodiversity. Findings reveal that mitigating both drivers significantly reduces extinction risk (−0.14), while unchecked climate change offsets gain from dam removal. The DSS highlights the dominance of climate stressors, with pollution and temperature shifts (−0.45, −0.42) as critical variables. Biological traits like reproductive frequency and longevity respond strongly to environmental improvements. This integrative approach bridges qualitative expertise and quantitative modeling, offering actionable insights for conservation planning. The study underscores the need for holistic strategies combining climate mitigation and habitat restoration to safeguard Mediterranean marine ecosystems. The FCM-based DSS provides a scalable tool for policymakers to prioritize interventions and assess trade-offs in complex socio-ecological systems.

1. Introduction

The Mediterranean Sea is a unique geographical region where Europe, Asia, and Africa converge, forming an exceptional ecological and cultural intersection [1]. Covering an area of approximately 2,500,000 square kilometers, it is the largest enclosed sea in the world and is home to an extraordinary variety of flora and fauna, shaped by the interaction of major continental landmasses [2]. With an average depth of 1500 m (4900 feet) and a maximum depth of 5109 m (16,762 feet) located in the Calypso Deep within the Ionian Sea, the Mediterranean exhibits notable bathymetric characteristics.

The sea supports one of the most diverse marine ecosystems globally. This biodiversity results from a combination of factors including its geographic location, which allows for the convergence of species from different regions, and its relatively stable temperature, salinity, and nutrient conditions. These characteristics foster a wide array of habitats, such as coastal zones, deep-sea trenches, and underwater caves, providing ecological niches for numerous marine organisms. Additionally, the Mediterranean’s complex geological evolution and its long-term isolation from other oceans have contributed to the emergence of many endemic species. However, this rich biodiversity is under increasing threat due to human activities including overfishing, habitat destruction, pollution, and climate change [1].

The Mediterranean is recognized as one of the most significant marine biodiversity hotspots, containing more than 17,000 marine species [1]. According to the IUCN Red List, approximately 20 percent of the assessed species in this region are classified as endangered [2]. Marine biodiversity is fundamental to maintaining ecological stability, resilience, and the overall health of ecosystems. It also supports critical economic activities such as fisheries, aquaculture, and tourism, while playing an essential role in preserving local cultures and traditions [3]. Mediterranean species provide indispensable ecosystem services. These include food supply, water and air purification, prevention of soil erosion, climate regulation, and pollination of crops by insects. The decline or loss of such species would have detrimental effects on both quality of life and economic stability. Therefore, their protection is not only a matter of environmental preservation but also a societal obligation to safeguard our natural heritage for the benefit of future generations [4].

Although Mediterranean countries have committed to conserving 10 percent of the sea by 2020 under the Convention on Biological Diversity (CBD), only about 4 percent of the area is currently protected, and just 0.01 percent is designated as a no-fishing zone [5]. The region remains vulnerable to a range of escalating threats, including global warming [6], the introduction of invasive species [7], excessive exploitation of marine resources, and other harmful human activities [8]. Addressing these challenges requires the expansion of spatially managed marine areas and the adoption of ecosystem-based approaches to both resource management [9] and marine spatial planning [10].

More than 1500 marine vertebrate and invertebrate species, including nearly one-third of all shark and ray species as well as reef corals and crustaceans, are currently facing substantial population declines and habitat degradation. These species are now considered at high risk of extinction [6]. Marine extinctions or the risk of extinction typically result from the interaction of multiple stressors [11]. These include harvesting activities combined with life-history and ecological traits that increase vulnerability, particularly in groups such as marine mammals, elasmobranchs, and sturgeons [11]. Species that are less resilient to anthropogenic pressures tend to display specific biological and ecological characteristics. For example, they often exhibit low reproductive rates, delayed maturity, rarity, narrow geographic distributions, and limited mobility during juvenile or adult stages [12]. Furthermore, such species may rely on habitats that are easily degraded by human activity, have specialized ecological requirements such as specific substrates or diets, or possess high commercial value that increases their risk of overexploitation, as is the case with sturgeons used for caviar production [11].

The Mediterranean Sea, a global biodiversity hotspot, is increasingly threatened by climate change, overfishing, pollution, habitat loss, and invasive species, placing numerous taxa at risk of extinction [1,2,6]. Despite international commitments, conservation remains fragmented, with only a small fraction of the region effectively protected [5]. Existing assessments often address threats in isolation and fail to capture the systemic interactions among ecological, technological, economic, and governance drivers of biodiversity loss [8]. This study addresses this gap by introducing an innovative decision-support framework that integrates a modified PESTEL analysis with Fuzzy Cognitive Mapping, enriched by expert knowledge and dimensionality reduction techniques. Unlike prior studies, this approach combines qualitative and quantitative tools to provide a holistic, scenario-based methodology for identifying critical drivers and prioritizing conservation interventions in the Mediterranean context.

The objective of this study is to investigate the key factors contributing to population declines among species in the Mediterranean Sea. By identifying major threats such as habitat loss, overfishing, pollution, and climate change, we seek to support the development of targeted conservation strategies aimed at mitigating these pressures. While earlier studies have demonstrated the usefulness of FCMs in environmental management [13,14], the present work is distinct. To address this gap, we propose a Decision Support System (DSS) that combines a modified BESTEL analysis, PCA-based dimensionality reduction, and FCM to provide a transparent and systematic approach for evaluating policy interventions. To our knowledge, no previous study has combined the above-mentioned models for this purpose. The integration of these tools allows us to construct scenario-based interventions and formulate conservation priorities that effectively address the most severe threats to endangered species. This approach offers a promising pathway for enhancing the protection of marine biodiversity in the Mediterranean region.

What makes this research particularly original, and worthy of publication is its novel methodological integration and applied focus on the Mediterranean Sea. By combining socio-political drivers (through the adapted PESTEL framework), statistical dimensionality reduction (via PCA), and expert-based modeling (using FCMs), this study bridges qualitative and quantitative domains in environmental assessment. Unlike previous studies that address threats in isolation, our approach provides a systems-level understanding of biodiversity loss, offering actionable insights through scenario analysis. This integrative methodology not only advances scientific knowledge but also delivers practical value to policymakers, conservation practitioners, and marine spatial planners aiming to protect one of the world’s most threatened and ecologically significant marine regions.

2. Materials and Methods

2.1. Development of BESTEL

Traditionally, PESTEL analysis, which includes political, economic, social, technological, legal, and environmental factors, has been employed as a strategic tool to evaluate the influence of external environments on organizations, particularly within corporate contexts. This framework was first introduced by Aguilar [15]. Each of these six factors, also referred to as dimensions, comprises key elements known as drivers that play a crucial role in shaping future strategic decisions and objectives [15]. Conducting a PESTEL analysis involves the collection of data from a wide range of sources, including academic literature, structured surveys, and other forms of information dissemination. The outcomes of this analysis are typically used to identify external threats and internal weaknesses that may affect an organization’s performance and adaptability (Professional Academy, n.d.).

The PESTEL framework has been widely applied in academic and policy-related research, including investigations into ocean energy development [16,17], the advancement of the Blue Economy [18], hydrogen production technologies [19], and strategies for regional development [20]. However, to the best of our knowledge, no prior studies have utilized the PESTEL framework to examine the effectiveness of ecological management strategies specifically aimed at marine conservation.

In response to this gap in the literature, we propose an adaptation of the original PESTEL model by replacing the political dimension with a biological dimension, thereby forming what we refer to as the BESTEL framework. This revised model includes biological, economic, social, technological, environmental, and legal factors (Figure 1). The primary objective of applying the BESTEL framework to the context of Mediterranean marine fauna is to analyze both internal and external variables that may influence the population dynamics and structural integrity of marine species within this ecologically sensitive region. The components of the BESTEL framework are detailed as follows:

• Biological factors encompass life history traits and biological characteristics of the marine species, such as the Phylum (Chordata, Mollusca, Arthropoda, Porifera), the mode of reproduction (Hermaphrodite, Gonochoristic, Asexually reproduction), the type of offspring (Oviparous, Viviparous, Ovoviviparous), the reproductive frequency within a year, the diet (Carnivore, Vegetarian, Omnivore), the age at maturity, the maximum age, the total body length, the competition and the migration.
• Economic factors (commercial value) relate to the fishing resource value provided by the exploitation of the species population.
• Social factors such as overfishing and poaching.
• Technological factors relate to the type of fishing gear, the scale of the fishing intensity as well as human related structures which may impact inland waterways and streams (dams).
• Environmental factors encompass the physical and ecological components of the proposed ecosystem. This includes aspects like pollution, climate change, non-indigenous species, temperature, depth, drought and eutrophication.
• Legal factors include spatial boundaries set between countries in general (living region) and by organizations, more specifically (Food & Agricultural Organization (FAO) Subareas)

This modification is intended to direct expert knowledge toward marine conservation topics that have not previously been systematically integrated, thereby offering practical insights to support conservation initiatives. An expert is generally defined as an individual with specialized knowledge in a specific domain, whose informed judgment is frequently relied upon for interpretation, analysis, and decision-making [21]. Such expertise may arise from formal education, extensive research, and the refinement of skills, but it can also be acquired through experiential learning and practical engagement in the field [22]. Expert opinion is commonly employed in the design and evaluation of research projects, particularly in critical phases such as hypothesis formulation, sample design, model development, and outcome interpretation [23,24,25]. In certain contexts, expert judgment may represent the only viable or most reliable source of information. This is particularly true when developing management strategies for modeling species distributions [26] or assessing the risk of colonization and spread of non-native species [27]. In such instances, expert insights serve as a foundational component of environmental decision-making frameworks [28]. By identifying the factors that experts regard as most influential in driving the extinction risk of the Mediterranean species, policymakers and researchers can develop more effective strategies to address emerging challenges. This understanding enables more proactive and adaptive planning, facilitating the alignment of conservation policies with ecological realities and future threats.

To achieve this, a questionnaire with a form of GDDP compliance was created and shared with experts through channels of institutionary communication. We elicited judgements of the level of influence of each factor from a diverse pool of 34 experts drawn from various fields: aquaculture, marine biology, ichthyology, and zoology. All participants involved in this study provided written informed consent. The role of the expert panel was to ascertain the relative importance of each criterion. To facilitate this, experts undertook a pairwise comparison for each criterion, using a 1–7 scale. The scale was defined as follows: 1 = low influence, 7 = high influence. At the end of the questionnaire, we asked the respondents to rate themselves from 1 (low) to 10 (high) how relevant to the subject they feel to use it to normalize the scores and give more value to answers from familiar with the problem experts. The equation that was used is the following:

$$\mathrm{n}\mathrm{o}\mathrm{r}\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{z}\mathrm{e}\mathrm{d} \mathrm{a}\mathrm{n}\mathrm{s}\mathrm{w}\mathrm{e}\mathrm{r}=\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{w}\mathrm{e}\mathrm{r}\times \mathrm{r}\mathrm{e}\mathrm{l}\mathrm{e}\mathrm{v}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{e}/10$$

(1)

Data questions span to all related BESTEL categories, and all respondents had to answer all questions regardless of their level of expertise. The questionnaire clearly defined the objectives of the research to ensure that all participants realize the level of question alignment with the research goals and provide relevant insights. Moreover, we guaranteed relevance and appropriateness of the questions to the target population and context, using language and terminology that are appropriate for the respondents. At the same time, simplicity and clarity were also kept since questions were clear, concise, and easy to understand to facilitate accurate responses and minimize confusion.

2.2. Principal Component Analysis

Principal Component Analysis (PCA) is a widely applied multivariate statistical method used to reduce the dimensionality of datasets containing interrelated variables while preserving the maximum amount of variance present in the original data. The core idea of PCA is to transform a set of possibly correlated variables $X=\{x_1,x_{2 },\dots x_n\}$ into a new set of linearly uncorrelated variables $Z=\{z_1,z_{2 },\dots z_n\}$, known as principal components. These components are ordered such that the first component $z_1$ captures the largest possible variance in the data, the second component $z_{2 }$ accounts for the next highest variance orthogonal to the first, and so on. Mathematically, PCA involves the eigenvalue decomposition of the covariance matrix $\mathrm{\Sigma }$ (or correlation matrix when variables are standardized), where

$$\sum = Q\Lambda Q^T$$

(2)

Here, $Q$ is the matrix of eigenvectors and $\Lambda$ is the diagonal matrix of eigenvalues. Each principal component is formed as a linear combination of the original variables:

$$z_i=q_{i1}{x}_1+q_{i2}{x}_2+\cdots +q_{in}{x}_n$$

(3)

where $q_{ij}$ denotes the loading of variable $x_j$ on the $i$-th principal component. The eigenvalues indicate the amount of variance explained by each principal component, allowing researchers to retain only those components that capture a significant portion of the total variance.

In the context of this study, PCA was applied to the BESTEL framework (Biological, Economic, Social, Technological, Environmental, and Legal factors) to extract the most influential variables shaping expert assessments of Mediterranean species decline. This transformed the original factor set into a smaller subset of orthogonal components and enabled the identification of latent structures and patterns among the dimensions. Moreover, this transformation reduced the original factor space by approximately 50 percent, allowing the model to focus on the most dominant contributors to systemic variability.

Importantly, the results of PCA provided the basis for the construction of the FCM. Since FCMs model complex systems through concepts (nodes) and causal relationships (edges), it is essential to select the most relevant variables for inclusion in the cognitive structure. The final concept set was derived through the structured PCA–expert integration protocol (Section 2.3) avoiding redundancy and enhancing interpretability. The reduced-dimensional space derived from PCA thus served as a foundation for defining the nodes and potential causal pathways in the FCM, ultimately supporting a more coherent and data-driven approach to modeling ecological and socio-environmental interactions affecting marine conservation.

2.3. PCA–Expert Integration Protocol

The integration of PCA outcomes with expert judgments followed a structured, two-stage protocol designed to ensure statistical rigor while preserving ecological and contextual relevance. This method avoids ad hoc inclusion or exclusion by applying explicit retention rules and a transparent expert validation process, documented in a decision audit trail. The procedure was adapted from established multivariate analysis and fuzzy cognitive modeling practices [29,30,31,32].

• Stage A—PCA Screening (Data-Driven Reduction)
• Step A1: Pre-processing.

All questionnaire-derived data were standardized (z-scores) to allow comparability across items [33]. Sampling adequacy was tested using the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy (>0.70 acceptable) and Bartlett’s test of sphericity (p < 0.001), confirming that the correlation matrix was suitable for PCA.

• Step A2: Component extraction.

Principal components were retained according to three complementary rules:

• Kaiser criterion: retain all components with eigenvalue λ ≥ 1.
• Scree test: retain all components before the “elbow” in the eigenvalue plot.
• Parallel analysis: retain components with eigenvalues greater than those from randomly permuted datasets [34].

The number of components was chosen such that cumulative explained variance exceeded 80%, balancing parsimony with explanatory power.

• Step A3: Variable retention rules.

Each variable (factor j) was evaluated with the following standardized criteria [30] as follows: (a) a variable j was provisionally retained if its maximum loading on any retained component was at least 0.40, (primary loading rule) (b) a variable j was flagged as ambiguous if the difference between its two highest loadings was <0.15 (Cross-loading rule) and (c) variables with communalities below 0.40 were flagged as weak contributors (Communality rule). Variables satisfying the primary loading rule without ambiguity were retained automatically. Variables failing one or more conditions were flagged for expert review in Stage B.

• Stage B—Expert Validation (Context-Driven Refinement)

Expert validation was then applied systematically to flagged variables to ensure that conceptually important factors were not excluded due to statistical thresholds.

• Step B1: Expert evaluation.

Flagged variables were re-evaluated by domain experts, who rated their indispensability on a 5-point Likert scale (1 = not relevant, 5 = highly indispensable). Experts were also invited to provide qualitative justifications (e.g., ecological necessity, policy importance).

• Step B2: Consensus criterion.

A flagged variable was retained if: P(retain) ≥ 0.70 and R⁻j ≥ 3.5 where P(retain) is the proportion of experts recommending retention, and R⁻j is the mean indispensability rating.

• Step B3: Theoretical override.

If consensus was not reached, a variable could still be retained if (a) its mean indispensability ≥ 3.5 and (b) at least two independent studies in the literature justified its inclusion (citations provided in Section 3).

2.4. Data Sources and Structure

The data used to construct and parameterize the model were collected via structured questionnaires administered to domain experts in aquaculture, marine biology, ichthyology, and zoology. The primary dataset consisted of 34 expert responses, each providing pairwise influence judgments among factors on a 1–7 scale (low → high influence). To account for differences in confidence and domain specialization, experts were also asked to provide a self-reported expertise score (1–10), which was later used as a weighting factor during data processing.

In addition, a supplementary survey was conducted with 37 students and academic professionals. This dataset was not used directly for FCM parameterization, but rather as a screening step to consolidate candidate concepts. Only items rated “somewhat relevant” or higher in this stage progressed to the expert round, ensuring that the expert dataset was focused on the most contextually meaningful factors.

The schema of the dataset maps directly to the BESTEL framework (see also Table 2 and Section 3 for the complete list and justification of factors). Each questionnaire produced:

• Nodes (concepts): the BESTEL factors, representing the components of the system.
• Edges (causal links): expert-assigned influence strengths between concepts, normalized to the range [−1…+1] and represented using a fuzzy triangular scale for uncertainty (e.g., “strongly important” → (5, 7, 9)).
• Meta-field: an expertise score (1–10) used for normalization and weighting.

This structure yielded three outputs: (i) a finalized list of concepts, (ii) a weighted, directed adjacency matrix representing causal relations, and (iii) respondent-level and aggregated metadata suitable for subsequent PCA-driven reduction and FCM construction.

In terms of size, the expert dataset of the 34 aforementioned completed questionnaires provided the pairwise influence assessments used for model building. The screening dataset consisted of 37 questionnaires, used only for initial filtering of concepts. Following PCA-guided reduction and expert validation, the final FCM model included 17 concepts and 70 directed connections (network density = 0.257).

For transparency, the anonymized questionnaire instrument, codebook, and aggregated datasets (weights, adjacency matrix, and derived inputs) are provided as Supplementary Materials (Dataset S1, Instrument S1).

2.5. Development of the FCM

FCMs are powerful modeling tools used to represent and analyze complex systems characterized by interdependent variables and feedback loops. Their strength lies in combining aspects of fuzzy logic and cognitive mapping to express systems in a form similar to human reasoning [35,36]. FCMs have been extensively applied in a wide range of disciplines, including crisis management and decision support systems [37], developmental planning [38], coastal management [14], and ecological modeling [39].

Technically, an FCM is a weighted directed graph in which nodes represent concepts (variables or factors), and edges represent weighted causal relationships among them. Each concept $C_i$ is a system component, and each directed edge $e_{ij}$ is associated with a weight $w_{ij} \in [-1, 1]$, indicating the strength and direction of the influence from concept $C_i$ to concept $C_j$ [33,34]. A positive weight indicates a causal increase, a negative weight indicates a causal decrease, and zero indicates no influence.

In an FCM, each concept represents a factor in the system (e.g., Climate Change, Pollution, Reproductive Frequency). The value of a concept reflects its current “activation level”, i.e., the intensity or state of that factor in the system. These values usually lie within a normalized range, such as [0, 1] or [−1, 1], depending on the transfer function used. More specifically: (a) a value close to 1 means the concept is highly active or strongly present (e.g., high pollution, strong climate change effect) and (b) a value close to 0 means weak activity or absence of the factor. In cases where negative values are used (e.g., with a hyperbolic tangent), negative values indicate inhibitory or adverse states. During simulation, these values are iteratively updated through the weighted causal relationships, until the system stabilizes. The final values of the concepts show how the whole system behaves under certain scenarios (e.g., with climate change mitigated vs. intensified).

To ensure interpretability and handle uncertainty, each concept in the FCM was expressed as a fuzzy set rather than a crisp variable. Fuzzy sets allow system components to be represented linguistically (e.g., Low, Medium, High) and mapped to numerical values through membership functions. In this study, a triangular fuzzy scale was used to express the intensity of importance between concepts, as shown in

Table 1. FCM State Analysis Metrics.

Definition	Crisp Values (Intensity of Importance)	$\mathbf{Fuzzy} \mathbf{Triangular} \mathbf{Scale} \tilde{\mathit{M}} = (\mathit{l},\mathit{m},\mathit{u})$
Equally important	1	(1, 1, 1)
Weakly important	3	(1, 3, 5)
Fairly important	5	(3, 5, 7)
Strongly important	7	(5, 7, 9)
Absolutely important	9	(7, 9, 9)

. This scale translates expert judgments into fuzzy numbers, capturing the inherent vagueness in assessing relationships. Each causal link between concepts in the FCM was assigned a fuzzy weight using this triangular scale. For example, if experts judged that Pollution strongly affects Water Quality, the relationship weight would be represented as the fuzzy number (5, 7, 9).

This approach ensures that expert uncertainty is systematically integrated into the model. The fuzzy weights were then processed through the FCM inference rules to update the concept values during simulation. Since decision-making requires crisp outputs, a defuzzification method was applied at the end of each simulation cycle. In this study, the Centroid (Center of Gravity) method was adopted, which computes the balance point of the aggregated membership function. This method was chosen because it is widely used in fuzzy cognitive modeling, produces stable outcomes, and provides interpretable crisp values.

Furthermore, to ensure reliability, consistency tests were performed on the FCM weights after de-fuzzification via Mental Modeler. Specifically, we examined whether the sign and direction of causal links matched expert expectations and conducted sensitivity tests by varying weights by ±10%. In addition, the number of retained principal components in the PCA step was altered to test stability. Across these checks, the ranking of the most influential drivers (climate change and dams) remained consistent, confirming the robustness of the model outcomes.

The logic of FCM functionality is based on forward iterations. This involves using mathematical formulation to calculate the values of the variables associated with each concept. In mathematical terms, an FCM with n concepts can be represented by a state vector ($S_n$) of length n that captures the values of each concept, and a Weight Matrix ($W_{n\times n}$) that captures the relationships among the concepts. Once the values ($S_i$) and weights have been assigned to the concepts ($C_i$), the FCM can converge to an equilibrium point after continuous forward iterations. The most popular inference rules are: (a) Kosko’s inference, (b) Modified Kosko’s inference and (c) Rescale inference as shown in the following three activation functions, respectively [40].

$$S_i\left(k+1\right)=f\left({\sum }_{j=1, j\ne i}^N{w}_{ji}\times S_j\left(k\right)\right)$$

(4)

$$S_i\left(k+1\right)=f\left(S_i\left(k\right)+\sum _{j=1, j\ne i}^N{w}_{ji}\times S_j\left(k\right)\right)$$

(5)

$$S_i\left(k+1\right)=f\left({(2\times S}_i\left(k\right)-1)+\sum _{j=1, j\ne i}^N{w}_{ji}\times {(2\times S}_j\left(k\right)-1)\right)$$

(6)

Also $f(.)$ is the threshold (transformation) function which can be (a) bivalent, (b) trivalent, (c) sigmoid or (d) hyperbolic, according to the following Equations (7)–(10), respectively:

$$f\left(x\right)=\left\{\begin{array}{c}1 x>0\\ 0 x\le 0\end{array}\right.$$

(7)

$$f\left(x\right)=\left\{\begin{array}{c}1 x>0\\ 0 x=0\\ -1 x<0\end{array}\right.$$

(8)

$$f\left(x\right)={\displaystyle \frac{1}{1+e^{-\lambda x}}}$$

(9)

$$f\left(x\right)=tanh(\lambda \times x)$$

(10)

with $\lambda$ to be a positive number $(\lambda >0)$ that establishes the continuous function’s $f$ steepness and $x$ is the value $S_i\left(k\right)$ on the equilibrium point. The sigmoid threshold function ensures that the estimated value of each concept will fall within the interval [0, 1]. Specifically, we must indicate that each concept $C_i$ has an associated activation value $A_i$∈[0, 1], which quantifies the degree to which the concept is “active” in the current system state. For example, a value close to 0 indicates minimal or no influence, a value near 0.5 indicates moderate activation, and a value near 1 indicates strong activation. These values evolve dynamically during simulation based on the weighted sum of incoming influences and the chosen threshold function. When the values of concepts are negative with values within the interval [−1, 1], we can use the hyperbolic tangent function instead of the sigmoid. In the structure of an FCM, concepts are further categorized into three types of nodes according to their role in the network:

• Driver nodes are those that primarily exert influence on other concepts but receive little or no influence or causalities themselves. They represent exogenous or initiating factors in the system.
• Receiver nodes are those that mainly accumulate influences or causalities from other nodes without significantly affecting others. They represent system outcomes or dependent variables.
• Ordinary nodes both influence and are influenced by other nodes. They represent intermediate factors that transmit effects through the system.

This classification enhances the interpretability of the FCM by clarifying which variables act as external drivers, which serve as mediating factors, and which capture the system’s final outcomes.

A graphical and matrix representation of a typical FCM is given in Figure 2.

There are two primary methods for building FCMs:

• Expert-driven construction: Concepts and causal weights are defined based on expert knowledge. This approach is particularly effective when empirical data are scarce or when human insight is crucial to understanding qualitative dependencies.
• Data-driven learning: When empirical data are available, FCM weights can be learned using algorithms such as Hebbian learning or nonlinear optimization methods to reduce error between predicted and actual concept [41,42]. In the present study, we employed an expert-driven FCM approach, informed by a preliminary Principal Component Analysis (PCA), to model the complex interactions influencing the decline of marine species in the Mediterranean Sea. The most significant factors identified through PCA were selected as FCM nodes. Expert judgment was then used to determine the nature and strength of causal relationships between these factors, forming the connection matrix W. The structure of the FCM not only enables the modeling of system behavior but also allows for scenario analysis and policy testing. By activating certain concepts and simulating changes in the system through iterative updates, stakeholders can evaluate the potential impacts of conservation measures and identify leverage points for intervention.

2.6. Development of a DSS for Marine Biodiversity Conservation Using FCM

In this study, a Decision Support System (DSS) was developed using an FCM to facilitate strategic planning and prioritization of conservation measures targeting endangered marine species in the Mediterranean Sea.

2.7. Decision Support System Architecture

The proposed Decision Support System (DSS) is designed as a modular, multi-stage framework that integrates qualitative and quantitative techniques to support evidence-based decision-making regarding marine species conservation in the Mediterranean Sea. The DSS consists of four interconnected layers:

• Input Layer—Data Collection and Driver Identification
  • The system begins with the modified BESTEL framework (an extension of PESTEL), which organizes the political, economic, social, technological, environmental, and legal factors influencing marine biodiversity.
  • Expert knowledge is systematically elicited through workshops and surveys, producing a comprehensive list of drivers related to species extinction.
  • The DSS accounts for both natural and anthropogenic pressures.
• Pre-processing Layer—Dimensionality Reduction
  • Complexity handler of the identified drivers via application of PCA.
  • PCA elimination of redundancies and grouping of correlated factors into principal components, thereby reducing noise and highlighting the most influential drivers.
  • Subsequent modeling focused on a manageable yet representative set of variables.
• Modeling Layer—FCM
  • Factors are translated to FCM concepts as nodes. Causal relationships are expressed as weighted edges.
  • The FCM structure captures both direct and indirect interactions among factors, enabling dynamic simulations of systemic behavior.
  • Stakeholders can assign and adjust weights based on expert consensus or scenario assumptions, making the model flexible and adaptive.
• Output Layer—Scenario Analysis and Decision Support
  • Via simulations on the FCM, the DSS evaluates the impacts of different interventions (e.g., stricter environmental regulations, reduction in dam-related pressures, climate adaptation policies).
  • The outputs are expressed in terms of system trajectories, showing how specific policy options influence the likelihood of mitigating species extinction.
  • Decision-makers may test several “what-if” scenarios and prioritize interventions with the greatest systemic impact.

Overall, the DSS provides a holistic and transparent framework that links qualitative insights with quantitative modeling. The proposed DSS offers a systematic framework to model the complex and often nonlinear interactions among biological, environmental, economic, technological, and legal factors that influence species decline. This system aims to support decision-makers in identifying critical variables, simulating policy scenarios, and evaluating their potential ecological impacts. The FCM was constructed by integrating expert knowledge and quantitative analysis. The initial list of relevant concepts was derived through questionnaires and opinions of experts. PCA was then applied to reduce the original concept set into a compact yet informative subset. These principal concepts were then mapped into nodes representing key concepts within the marine socio-ecological system. Directed edges between nodes were established to reflect causal relationships, with associated weights expressing the strength and direction of influence, as determined through structured expert elicitation.

The resulting cognitive model is represented by a weighted adjacency matrix $W$ where each element $w_{ij}\in [-1, 1]$ denotes the causal impact of concept $C_i$ on concept $C_j$. The DSS employs iterative computations to simulate the propagation of change through the system. Specifically, the system dynamics are captured by:

$$A_i^{t+1}=f(A_i^t+ {\sum }_{j\ne i}{A}_j^t.w_{ji})$$

(11)

where $A_i^t$ is the activation level of concept $C_i$ at iteration t, and f(⋅) is a transfer function that bounds the output within [0, 1]. In this work, a sigmoid function was employed to preserve sensitivity to incremental changes while ensuring stability in the simulation results.

Several scenario analyses were performed to explore potential outcomes of different policy interventions and environmental changes. By altering the initial activation levels of selected concepts (Trawling, Small Case Fisheries, Poaching, Climate Change, Type of Offspring), the DSS simulated downstream effects on Extinction of Sea Life, Fishing Revenues, Biodiversity, species extinction risk, and overall ecosystem health. These simulations offer insights into which factors exert the most influence over the system and which leverage points may yield the greatest conservation benefits.

Moreover, the DSS facilitates participatory planning by making system dynamics transparent and comprehensible to multiple stakeholders, including marine ecologists, conservation agencies, local communities, and policymakers. The engagement of domain experts in the design and validation of the cognitive structure promotes the system shared understanding, consensus-building, and informed decision-making.

3. Development and Justification of the BESTEL Framework for Assessing Marine Species Extinction Risk

This adaptation enables a tailored, ecologically meaningful categorization of determinants that are specifically relevant to understanding biodiversity loss in marine ecosystems.

Table 2. List of all BESTEL factors affecting the extinction of sea life in the Mediterranean Sea [43].

Category	Factors	References
Biological	Phylum	[6,44,45]
	Mode of reproduction	[46,47]
	Type of offspring	[48,49,50]
	Number of offspring	[51,52,53,54]
	Reproductive frequency	[55,56,57]
	Diet	[55,58,59,60,61]
	Age at maturity	[55,62,63,64]
	Maximum age	[56,65]
	Competition	[66]
	Total body length	[67,68]
	Migration	[57,69,70,71]
Economic	Commercial value	[60,72,73,74]
Social	Overfishing	[66,75]
	Poaching	[76,77]
Technological	Small scale fisheries	[67]
	Trawling	[78,79]
	Purse seine	[55,80]
	By-catch	[74]
	Dams	[64,81]
Environmental	Pollution	[72,82,83]
	Eutrophication	[69,84]
	Drought	[73,85]
	Depth	[62,67,86]
	Temperature	[58,64,68]
	Climate Change	[58,67,81,87]
	Nonindigenous Species	[51,88,89]
Legal	Living region	[59,65,90]

outlines the full list of 27 factors classified under the six BESTEL domains, along with their corresponding literature references. The following section provides a detailed justification for the inclusion of each factor in the analysis.

• Biological Factors: Biological traits are often the most immediate predictors of a species’ vulnerability to extinction. We included 11 biological variables:

• Phylum [6,44,45]: Taxonomic classification is essential as phylogenetic relatedness is associated with ecological traits and extinction risk. Certain phyla may be more susceptible to environmental changes or human exploitation due to physiological constraints or reproductive modes.
• Mode of reproduction [46,47]: Species that reproduce sexually, asexually, or through parthenogenesis exhibit differing levels of genetic variability and population resilience, impacting their capacity to adapt to changing environments.
• Type of offspring [48,49,50]: Species with planktonic larvae may disperse widely and colonize new areas, whereas species with brooded or direct-developing offspring are more site-attached and potentially more vulnerable to habitat changes.
• Number of offspring [51,52,53,54] and Reproductive frequency [55,56,57]: These traits affect population recovery rates. Species with low fecundity and infrequent reproduction tend to be less resilient to anthropogenic pressures.
• Diet [55,58,59,60,61]: Trophic specialization often correlates with extinction risk. Species with narrow dietary requirements are more vulnerable to food web disruptions.
• Age at maturity [55,62,63,64] and Maximum age [56,65]: Late-maturing and long-lived species generally have slower life histories, making them more susceptible to overexploitation.
• Competition [66]: Inter- or intraspecific competition for limited resources can reduce population sizes and increase vulnerability, especially under environmental stress.
• Total body length [67,68]: Larger-bodied marine species tend to be preferentially targeted by fisheries and also have slower reproductive rates.
• Migration [57,69,70,71]: Migratory species require multiple habitats across their life cycles, making them particularly sensitive to habitat fragmentation, barriers, and climatic changes.

• Economic Factors

• Commercial value [60,72,73,74]: Market demand exerts direct pressure on species. Highly valuable species are more heavily exploited, often unsustainably. Their inclusion helps quantify the economic incentive driving extraction.

• Social Factors

• Overfishing [66,75]: One of the most pervasive threats to marine biodiversity, overfishing reduces population sizes below sustainable levels and alters food web dynamics.
• Poaching [76,77]: Illegal, unreported, and unregulated (IUU) fishing exacerbates population decline, especially for endangered or protected species, undermining formal conservation efforts.

• Technological Factors: Technological advances have enhanced fishing efficiency but also increased the ecological footprint of marine resource exploitation.

• Small-scale fisheries [67]: Though generally seen as less destructive, these operations can cause localized overexploitation in coastal areas where many species reproduce.
• Trawling [78,79]: A non-selective and habitat-damaging method known to cause significant seabed disruption and high by-catch rates.
• Purse seine [55,80]: Commonly used for pelagic species, it can contribute to population collapse if poorly managed.
• By-catch [74]: Incidental capture of non-target species remains a significant source of unmonitored mortality, particularly for vulnerable or rare taxa.
• Dams [64,81]: By altering water flow and fragmenting habitats, dams disrupt critical ecological processes, particularly for diadromous or migratory species.

• Environmental Factors: Marine species are directly affected by abiotic conditions and broader ecological dynamics. The following factors were included for their empirical association with biodiversity change:

• Pollution [72,82,83]: Introduces toxins and disrupts biochemical cycles, directly harming marine life and degrading habitats.
• Eutrophication [69,84]: Nutrient overloading causes hypoxia and harmful algal blooms, leading to habitat degradation and mortality.
• Drought [73,85]: Reduces freshwater inputs into estuaries and coastal systems, altering salinity and nutrient balances.
• Depth [62,67,86]: Species exhibit vertical distribution preferences; bathymetric range may influence susceptibility to fishing or environmental fluctuations.
• Temperature [58,64,68]: Affects metabolic rates, reproduction, and distribution. Sudden or chronic increases are particularly threatening to stenothermic species.
• Climate Change [58,67,81,87]: Represents a systemic driver impacting multiple environmental variables, from sea level rise to acidification.
• Nonindigenous Species (NIS) [51,88,89]: Invasive species may outcompete, predate upon, or introduce diseases to native species, disrupting native ecosystems.

• Legal Factors

• Living region [59,65,90]: Species’ exposure to legal protection varies across jurisdictions and management regimes. Legal status and protected area coverage influence exploitation rates and conservation outcomes.

The selection of these factors was based on both theoretical ecological models and empirical findings from literature. Their inclusion allows the BESTEL framework to capture a wide range of pressures that affect marine biodiversity, from intrinsic species traits to socio-economic drivers and governance structures. Furthermore, these variables provided the foundational input for the subsequent Principal Component Analysis (PCA) and the development of the FCM, ensuring that the model reflects the real-world complexity and interdependence of marine species extinction dynamics in the Mediterranean region.

4. Results and Scenario Analysis

PCA was used to condense this multidimensional data into a smaller set of principal components, retaining the most significant variance in the original dataset while reducing noise and redundancy. PCA identified the dominant factors/dimensions driving variability within the BESTEL framework. After we examined the loadings of variables, we discerned which aspects of the external environment have the most substantial impact on the organization or system under analysis.

4.1. Case Study Description: Living Region of Mediterranean Marine Species

The present study focuses on the spatial distribution of marine species at risk of extinction in the Mediterranean Sea, with a particular emphasis on their living region. This factor, categorized under the Legal dimension of the BESTEL framework, serves as a crucial variable in understanding exposure to spatially dependent anthropogenic pressures, regulatory protection, and habitat-specific stressors. To operationalize this variable, the Food and Agriculture Organization (FAO) statistical divisions for the Mediterranean Sea were adopted as the geographic reference framework. These FAO subareas are internationally recognized zones used for fishery and biodiversity assessments, enabling spatially explicit analysis and facilitating alignment with conservation policy and marine spatial planning (Figure 3).

Each of these regions varies considerably in terms of environmental characteristics (e.g., depth gradients, temperature regimes, salinity), human activities (e.g., fishing intensity, pollution levels), and degree of legal protection (e.g., marine protected areas, fishing restrictions).

• Spatial Distribution of Threatened Species

Species occurrence and extinction risk data were spatially mapped to the corresponding FAO subareas, using curated records from the IUCN Red List, regional assessments, and published literature [59,65,90]. A heatmap was generated to visualize the concentration of endangered species across the Mediterranean subregions (see Figure 3). The intensity of the color scale represents the relative species richness and extinction threat level, with warmer tones indicating higher density and vulnerability.

• Justification for the Living Region as a Legal Factor

The inclusion of the living region as a Legal factor within the BESTEL framework is justified on the grounds that the regulatory framework and enforcement capacity vary significantly across FAO subareas. While some regions are embedded within marine protected areas (MPAs) with active enforcement and ecological monitoring, others lack adequate conservation infrastructure, making them more susceptible to unregulated exploitation and habitat degradation. Furthermore, the designation of FAO subareas allows for a standardized comparative analysis across political and ecological boundaries, and it facilitates policy-relevant recommendations that align with international conventions such as the Barcelona Convention and the EU Marine Strategy Framework Directive.

4.2. Dimensionality Reduction in BESTEL Factors Using Principal Component Analysis (PCA)

Given the multidimensional structure of the BESTEL framework, which includes a total of 27 factors across six domains, PCA was employed to condense the number of variables, focusing particularly on the Biological and Environmental dimensions, which contained the largest number of variables (11 and 7, respectively). PCA transforms the original set of correlated variables into a new set of orthogonal variables called principal components (PCs), which are linear combinations of the original factors. These components are ordered by the amount of variance they capture from the data. The mathematical formulation of PCA involves computing the eigenvalues and eigenvectors of the correlation matrix $R$ of the standardized data matrix $X$, where

$$R={\displaystyle \frac{1}{n-1}} X^T$$

(12)

with n = 34 to be the number of expert questionnaires.

Each principal component ${PC}_k$ is defined as ${PC}_k=a_{k1}{x}_1+a_{k2}{x}_2+\cdots +a_{kn}{x}_n$.

• Application to Biological Factors

To demonstrate the contribution of PCA to factor selection, we applied the analysis specifically to the biological subset of factors from the expert dataset. The input data consisted of 34 expert questionnaires as mentioned before, each providing pairwise influence judgments among the BESTEL framework variables. From these, the biological factors adjacency sub-matrix was extracted, comprising Mode of Reproduction, Number of Offspring (Fecundity), Reproductive Frequency, Age at Maturity, and Maximum Age (Longevity). The PCA was conducted on the correlation matrix of this sub-matrix, as the variables were standardized prior to analysis. Sampling adequacy was verified with KMO = 0.74 and Bartlett’s test (p < 0.001), confirming the suitability of PCA. The analysis yielded two principal components with eigenvalues > 1, together explaining 82% of the variance.

The component loadings (

Table 3. PCA Loadings for Biological Factors.

Factor	PC1 (Reproductive Capacity)	PC2 (Life-History Longevity)	$\mathbf{Communality} \left({\mathit{h}}^2\right)$
Phylum	0.21	0.18	0.08
Mode of Reproduction	0.782	0.321	0.72
Type of Offspring	0.27	0.22	0.12
Number of Offspring	0.822	0.284	0.75
Reproductive Frequency	0.768	0.357	0.71
Diet	0.33	0.29	0.19
Age at Maturity	0.291	0.815	0.75
Maximum Age (Longevity)	0.277	0.846	0.78
Competition	0.36	0.24	0.18
Total Body Length	0.39	0.27	0.22
Migration	0.34	0.31	0.21

) showed that all five factors contributed significantly (>0.60) to at least one retained component. Specifically, Mode of Reproduction, Number of Offspring, and Reproductive Frequency loaded strongly on the first component, reflecting reproductive capacity, while Age at Maturity and Maximum Age (Longevity) loaded primarily on the second component, reflecting life-history timing and survival potential. This result validates that the selected biological factors capture two distinct but complementary dimensions of species dynamics: (i) reproductive capacity and (ii) life-history longevity. Together, they provide a balanced representation of biological influences on population sustainability.

To further illustrate these findings, Figure 4 presents a PCA biplot of the five biological factors. The clustering pattern confirms the grouping of Mode of Reproduction, Number of Offspring, and Reproductive Frequency under one latent dimension, and Age at Maturity and Maximum Age under another.

• Application to Environmental Factors

To validate the selection of environmental factors, we applied again PCA to the expert dataset restricted to environmental-related variables: Pollution, Eutrophication, Drought, Depth, Temperature, Climate Change, and Non-indigenous Species. The input data consisted of the aggregated and normalized adjacency sub-matrix (34 expert questionnaires), previo0usly used. The same PCA criteria used (i.e., sampling adequacy was verified with KMO = 0.71 and Bartlett’s test of sphericity (p < 0.001), confirming the suitability of PCA). Two principal components with eigenvalues greater than 1 were extracted, explaining 80.6% of the total variance.

The component loadings in

Table 4. PCA Loadings for Environmental Factors.

Factor	PC1 (Anthropogenic)	PC2 (Climatic)	$\mathbf{Communality} \left({\mathit{h}}^2\right)$
Pollution	0.824	0.254	0.74
Eutrophication	0.453	0.326	0.30
Drought	0.216	0.584	0.38
Depth	0.185	0.623	0.42
Temperature	0.277	0.797	0.71
Climate Change	0.321	0.839	0.78
Non-indigenous species	0.845	0.265	0.74

show that Pollution, Temperature, Non-indigenous Species and Climate Change load strongly (>0.70) on the retained components, indicating their dominant contribution to explaining variance in the environmental dataset. By contrast, variables such as Eutrophication, Depth and Drought exhibited weaker or cross-loadings, and lower communalities, suggesting redundancy or insufficient explanatory contribution.

This analysis demonstrates that the four retained factors represent two complementary environmental dimensions: (i) anthropogenic stressors (Pollution) and (ii) climatic drivers (Temperature and Climate Change). Together, they capture the majority of the explanatory power of the environmental subset while avoiding redundancy.

The grouping pattern is visualized in Figure 5, which presents the PCA biplot of the seven environmental variables. The biplot confirms the clustering of Pollution, Temperature, and Climate Change as primary drivers along the first two components, while the remaining variables contribute less consistently. Thus, PCA combined with expert validation provides a robust justification for selecting Pollution, Temperature, and Climate Change as the key environmental factors included in the final FCM model.

• Application to Technological Factors

To evaluate the contribution of technological pressures, PCA was applied also to the expert dataset restricted to the technological variables: Small-scale Fisheries, Trawling, Purse Seine, By-catch, and Dams. The input was the same as above. PCA was performed on the correlation matrix of these factors (standardized), with KMO = 0.72 and Bartlett’s test (p < 0.001) confirming sampling adequacy. Two principal components with eigenvalues > 1 were retained, explaining 81.4% of the total variance.

The component loadings in

Table 5. PCA Loadings for Technological Factors.

Factor	PC1 (Fishing Practices)	PC2 (Infrastructure)	$\mathbf{Communality} \left({\mathit{h}}^2\right)$
Small-scale Fisheries	0.42	0.31	0.27
Trawling	0.84	0.22	0.76
Purse Seine	0.48	0.36	0.35
By-catch	0.81	0.28	0.74
Dams	0.27	0.85	0.80

indicate that Trawling and By-catch load strongly on the first component, representing fishing practice intensity, while Dams loads primarily on the second component, representing infrastructure-related environmental alteration. In contrast, Small-scale Fisheries and Purse Seine exhibited weaker loadings (<0.50) and lower communalities, suggesting lower explanatory contribution and redundancy with the retained factors.

This outcome was further confirmed by the PCA biplot (Figure 6), which shows the clustering of Trawling and By-catch along one axis and Dams along another, clearly distinguishing them as the dominant technological drivers. Consequently, the selection of Dams, Trawling, and By-catch is both empirically supported by PCA and theoretically justified as they capture the main technological pressures affecting marine ecosystems.

• Application to Economic, Social and Legal Factors

The economic dimension was represented by the factor Commercial Value. In this category, only one variable was elicited during the expert questionnaire round, reflecting the consensus that market valuation is the dominant economic driver of species vulnerability. Because the set contained only a single factor, no PCA-based dimensionality reduction was needed. Instead, the inclusion of Commercial Value was confirmed directly by experts, who rated it as indispensable for capturing the role of economic incentives in driving exploitation pressures. This factor was therefore retained without further screening.

The social dimension included two factors: Overfishing and Poaching. Both were consistently highlighted during the expert elicitation phase as primary social pressures influencing marine species sustainability. As the category contained only two items, PCA was not applicable. Instead, the decision to retain both factors was based on expert consensus (retention > 90%) and their theoretical centrality: Overfishing represents widespread unsustainable harvesting practices, while Poaching reflects illegal and unregulated activities. Together, they captured the spectrum of social exploitation pressures and were therefore both retained.

The legal dimension was represented solely by Living Region, reflecting jurisdictional and regulatory constraints on species exploitation. This factor was suggested during the initial factor elicitation stage and validated by the expert panel as an indispensable component of the BESTEL framework. Given that it is the only legal factor, PCA-based reduction was not relevant. Its inclusion in the final model is therefore both necessary and sufficient to represent the legal-regulatory context influencing species vulnerability.

4.3. FCM Implementation

The development of the FCM was carried out through a structured, step-by-step process. Initially, a comprehensive literature review was conducted to identify key factors contributing to the extinction of marine life in the Mediterranean region. The identified concepts were systematically classified into six overarching categories, corresponding to the domains outlined in the previously introduced BESTEL framework. To validate and enhance this classification, an additional survey was administered to academic experts, who were also invited to suggest supplementary variables or concepts where appropriate. At the same time, these experts evaluated the importance of each concept by assigning significance weights.

In the second phase, a questionnaire was distributed to students and academic professionals specializing in ichthyology across two Greek universities. A total of 37 completed questionnaires were collected. Responses were analyzed, and only those concepts rated as “somewhat relevant” or higher were retained for further review by the experts. Suggestions for new concepts were evaluated, and similar or overlapping concepts were consolidated to streamline the final FCM structure. Subsequently, a refinement stage was undertaken by two subject matter experts. This stage incorporated insights from the earlier PCA and ensured that all BESTEL domains were represented in the final concept map. This led to a condensed and more functional version of the FCM comprising 17 total concepts. Of these, four were classified as receiver variables (pollution, temperature, living region, and extinction of sea life), while two were identified as drivers (climate change and dams).

Experts were then asked to assign weights to the causal relationships among the concepts using a scale from −1 to +1. To model the intensity of these causal influences, the hyperbolic tangent (tanhx) function was selected as the squashing function, due to its ability to effectively capture nonlinear interactions. Prior to conducting steady-state analysis or simulating scenarios, a detailed understanding of the dynamics and structure of the system was established.

The construction of the FCM was carried out with structured input from domain experts to ensure the inclusion of all relevant concepts and the correct specification of their interrelationships. To model the causal strengths, we employed fuzzy set theory, which allows linguistic assessments to be translated into quantitative weights while explicitly accounting for uncertainty and vagueness in expert judgments. Specifically, each expert evaluated the causal influence between pairs of concepts using a seven-point linguistic scale (e.g., Very Low, Low, Medium, High, Very High and their negative counterparts for inverse effects). Each linguistic label was formally represented by a triangular fuzzy number $\tilde{\mathrm{M}}=(\mathrm{l},\mathrm{m},\mathrm{u})$, where l, m, and u denote the lower, modal, and upper bounds of the membership function, respectively (see Table 1). For example, the term Strongly Important corresponds to the fuzzy number (5, 7, 9), reflecting the possibility that the true weight lies anywhere in that range, with 7 as the most plausible value. This procedure transforms qualitative linguistic descriptors into fuzzy sets that encode expert uncertainty rather than reducing them to crisp values prematurely. The aggregation of multiple expert opinions was performed through the union of fuzzy sets, yielding a consensus fuzzy weight for each causal connection. To incorporate these into the FCM simulation, the fuzzy weights were subsequently de-fuzzified into crisp numerical values. We adopted the widely used Centroid (Center of Gravity) method, which calculates the balance point of the membership function and yields stable, interpretable outputs. Thus, the final FCM weight matrix W consisted of crisp values in the interval [−1, 1], derived systematically from fuzzy linguistic inputs. This approach ensured that the uncertainty inherent in human judgment was preserved during the modeling process and only resolved at the final stage when crisp weights were required for simulation.

The concept nodes were derived as follows: five from the PCA of biological factors, four from the PCA of environmental factors, and three from the PCA of technological factors. In addition, four nodes were taken directly from the social, economic, and legal categories, and one final node (Extinction of Sea Life) was included as the primary variable of interest in our study. This last concept also functions as one of the four receiver nodes in the FCM, serving as the reference point against which all scenarios were evaluated.

These qualitative assessments were then de-fuzzified using the Centroid (Center of Gravity) method within the Mental Modeler platform (http://www.mentalmodeler.com, accessed on 21 June 2025) and integrated into the FCM model, as visualized in Figure 7. It is important to note that the weights shown on the map reflect average de-fuzzified opinions and are subjective estimations rather than empirically measured values. Respondents were instructed to evaluate each concept’s potential influence on all other concepts by assigning values representing the perceived strength and direction of the relationship. This comprehensive approach allowed for the calculation of key structural metrics such as in-degree, out-degree, and centrality for each receiver concept in the map.

Table 6. FCM State Analysis Metrics.

Component	Indegree	Outdegree	Centrality	Type
Mode of Reproduction	2.489	1.589	4.08	ordinary
Number of Offspring	2.760	3.219	5.98	ordinary
Reproductive Frequency	2.14	2.949	5.09	ordinary
Age at Maturity	2.42	3.839	6.26	ordinary
Maximum Age (Longevity)	2.48	4.02	6.5	ordinary
Commercial value	2.859	1.65	4.51	ordinary
Overfishing	2.88	3	5.88	ordinary
Poaching	0.909	1.720	2.63	ordinary
Trawling	0.72	2.58	3.3	ordinary
By-catch	1.98	1.81	3.79	ordinary
Dams	0	1.909	1.909	driver
Climate Change	0	4.42	4.42	driver
Nonindigenous Species	0.89	2.29	3.18	ordinary
Pollution	1.22	0	1.22	receiver
Temperature	1.31	0	1.31	receiver
Living region	1.52	0	1.52	receiver
Extinction of Sea Life	8.42	0	8.42	receiver

presents the results of the steady-state analysis of the FCM, for which calculations gave Total Components = 17, Total Connections = 70, Density = 0.257, Connections per Component = 4.117, Number of Driver Components = 2, Number of Receiver Components = 4, Number of Ordinary Components = 11, Complexity Score = 2.234 with specific statistics depicted in this table.

4.4. Scenario Analysis

To ensure transparency and reproducibility, the initial activation values of all concepts were explicitly defined before running the scenario simulations. These values represent the baseline state of the system, derived from the steady-state outputs of the FCM model following convergence. Each concept was normalized within the range [0, 1], where 0 indicates absence or minimal influence, 0.5 represents moderate activation, and 1 corresponds to maximum activation. During scenario simulations, only the designated driver nodes (e.g., Climate Change and Dams) were perturbed to test alternative futures. All other concepts retained their baseline values. Deviations reported in the results are therefore relative to this initial state, ensuring comparability across scenarios. The baseline activation values used to initialize the scenario simulations are presented in the following

Table 7. Baseline Initial Values of Concepts Used in Scenario Simulations.

Concept	Initial Value (0–1)	Source-Justification
Mode of Reproduction	0.50	Expert elicitation; equilibrium state
Number of Offspring	0.55	Expert elicitation
Reproductive Frequency	0.50	Derived from FCM equilibrium
Age at Maturity	0.45	Expert elicitation
Maximum Age (Longevity)	0.40	Expert elicitation
Commercial value	0.55	Expert elicitation
Overfishing	0.60	Expert panel consensus
Poaching	0.35	Expert elicitation
Trawling	0.50	Expert elicitation
By-catch	0.45	Expert elicitation
Dams	−1 or 1 (worst or best)	Expert elicitation (driver value)
Climate Change	−1 or 1 (worst or best)	Expert elicitation (driver value)
Nonindigenous Species	0.35	Expert elicitation
Pollution	0.30	Expert elicitation
Temperature	0.50	Derived from FCM equilibrium
Living region	0.45	Expert elicitation
Extinction of Sea Life	0.40	Expert elicitation

.

We implemented a comprehensive scenario analysis to assess the systemic response of Mediterranean marine ecosystems under varying configurations of anthropogenic stressors, using the FCM model as the analytical backbone. This process allowed us to explore how different combinations of the two primary driver concepts (Climate Change and Dams) influence the trajectory of marine biodiversity and extinction risk. Four policy-relevant scenarios were designed, each reflecting plausible socioenvironmental futures:

• Mitigating Structural Pressures (Climate Change = −1, Dams = −1), representing an ideal condition with effective climate action and infrastructure decommissioning.
• Climate Intensification with Dam Mitigation (Climate Change = +1, Dams = −1), depicting a situation where climate threats escalate despite improved riverine connectivity.
• Dual Stress Intensification (Climate Change = +1, Dams = +1), simulating a worst-case scenario with compounded global and regional pressures.
• Climate Mitigation with Continued Damming (Climate Change = −1, Dams = +1), testing the extent to which climate action alone can buffer against persistent structural barriers.

For each case, the sigmoid activation function was employed to ensure convergence and interpretability of system behavior. In all simulations, concept values were updated iteratively until the system reached a stable equilibrium state. Convergence was typically achieved within 20–25 iterations across all scenarios. To ensure robustness, we adopted a fixed maximum of 50 iterations, consistent with standard practices in FCM applications [30,34]. This approach guarantees that transient fluctuations are resolved and that the final concept activation values represent the steady-state behavior of the system under each scenario.

The initial conditions were based on the steady state outputs of the FCM, and deviations in concept activations were analyzed post-simulation. Special focus was placed on the receiver nodes (Extinction of Sea Life, Pollution, Temperature, and Living Region) as key indicators of ecosystem health. This analytical framework enables the identification of nonlinear feedback, synergies, and trade-offs between structural and climatic interventions, providing valuable insights into the design of integrated marine conservation and policy strategies. Detailed interpretations of each scenario’s outcomes are provided in the subsections that follow.

4.4.1. Scenario: Mitigating Structural Pressures (Climate Change = −1, Dams = −1)

This scenario investigates the effects of mitigating two major anthropogenic drivers: Climate Change and Dams, both set to their minimum activation level (−1), simulating ideal policy implementation and technological solutions that alleviate their ecological burden. The simulation output is illustrated in Figure 8 (uploaded image), which displays the change in state values of all concepts in the FCM model. The key findings of this simulation are the following:

• The Extinction of Sea Life, which is the central outcome node, exhibits a negative variation (−0.14), indicating a significant reduction in extinction risk. This demonstrates that targeting structural ecosystem stressors can alleviate cumulative pressures on marine biodiversity.
• The Temperature and Pollution concepts show the largest negative shifts (−0.45 and −0.42, respectively). This is consistent with the ecological literature indicating that mitigating greenhouse gas emissions and restoring freshwater connectivity reduces ocean warming, acidification, and chemical flux into marine ecosystems.
• The nonindigenous species concept also registers a moderate decline (−0.09), reflecting the ecological benefits of stabilizing climate systems and hydrological flows. Warmer, disturbed environments often promote biological invasions; thus, reducing stress improves ecological resistance.
• Interestingly, several biological traits (e.g., Reproductive Frequency, Number of Offspring, and Maximum Age) exhibit positive increases in activation (0.14–0.15), indicating improved biological resilience. These species-level traits are indirectly bolstered by a more favorable environmental baseline (e.g., improved temperature ranges, reduced stress on reproductive cycles).
• The mode of reproduction concept slightly decreases (−0.09), possibly due to the interplay between environmental stability and reproductive strategy (e.g., broadcast spawners may experience reduced reproductive output under cooler, less turbulent waters), but the effect is marginal.
• The Living Region, a legal/geographical proxy, increases in value (0.20), suggesting that mitigation of large-scale stressors enhances the carrying capacity and suitability of habitats for diverse marine species, particularly in sensitive FAO subregions of the Mediterranean.
• Finaly, the technological concepts like trawling and by-catch remain largely unaffected, reinforcing that technological policies require direct intervention rather than benefiting passively from climate or hydrological improvements.

For this scenario we conclude that it highlights the amplified benefits of systemic environmental restoration. Reducing climate change and removing migration barriers like dams leads to multiscale improvements from organismal traits to ecosystem-level dynamics, ultimately lowering the extinction trajectory. The simulation confirms that upstream ecological interventions have downstream biodiversity dividends and should be considered priority actions in Mediterranean marine conservation policy.

4.4.2. Scenario: Intensification of Climate Change and Mitigation of Dams (Climate Change = 1, Dams = −1)

This scenario models a future in which climate change impacts are exacerbated (Climate Change = +1), while mitigation efforts have successfully removed or reduced the impact of dams, (Dams = −1). This mixed policy condition reflects a plausible trajectory in the Mediterranean where progress is made in freshwater connectivity, but global climate mitigation lags. The results are visualized in the uploaded Figure 9 and described below:

• The Extinction of Sea Life, which is the central outcome of the FCM, shows a slight increase (+0.01) despite dam mitigation. This suggests that climate change alone can offset gains made in other sectors, affirming its dominant role as a structural ecological stressor.
• Several biological traits worsen in their predicted state:

  ○

  Mode of Reproduction shows a notable negative shift (−0.10).

  ○

  Number of Offspring (−0.06) and Reproductive Frequency (−0.06) also decline.

  ○

  These changes reflect how climate stressors such as temperature rise, ocean acidification, and hypoxia disrupt species’ reproductive strategies and success, particularly in low-fecundity or late-maturing species.

• Longevity and Age at Maturity decrease slightly (−0.03), aligned with the biological consequences of accelerated metabolic rates, reduced survival, and habitat stress under warming seas.
• The Living Region concept shows a strong positive increase (+0.20), indicating that dam mitigation enhances habitat accessibility and ecological suitability. This improvement reflects better migratory conditions for species like eels and salmon, which depend on river-ocean connectivity.
• Despite the positive shift in Living Region, other environmental stressors degrade:

  ○

  Nonindigenous Species (−0.10) increase in pressure, likely due to warmer waters favoring invasive species.

  ○

  Pollution (−0.09) and Temperature (−0.01) also exhibit deterioration, driven by climate intensification which worsens bioaccumulation, eutrophication, and habitat stress.

• Notably, commercial, technological, and social pressures (e.g., Overfishing, Poaching, Trawling) remain stable, showing minimal variation. This highlights that unless directly targeted, such sectors do not respond to indirect changes in climate or hydrological regimes.

This scenario highlights a critical insight: mitigating one domain (e.g., dams) is not enough to counterbalance the ecological harm driven by unchecked climate change. While dam removal enhances migratory routes and habitat viability (as shown by the boost in Living Region), the reproductive traits of marine species and the intensity of invasion and pollution still deteriorate, leading to a net stagnation in extinction risk. Strategic conservation must therefore prioritize climate adaptation and resilience, not just infrastructure removal, to prevent biodiversity loss in the Mediterranean.

4.4.3. Scenario: Intensification of Climate Change Combined with Proliferation of Dams (Climate Change = +1, Dams = +1)

This scenario explores a worst-case trajectory where both climate change impacts are intensified, and dams increase. This reflects a plausible yet ecologically hazardous pathway for the Mediterranean and surrounding basins, where climate policies stall and freshwater-marine systems are further fragmented by infrastructural expansion. Observing the simulations results depicted in Figure 10 we describe and explain these findings:

• The Extinction of Sea Life shows a positive increase (+0.02), indicating an elevated extinction risk in marine biodiversity under dual pressure. While the increase is relatively modest, it signifies the compounding nature of multiple anthropogenic stressors.
• The Living Region concept exhibits a marked decline (−0.05). This reinforces the disruptive role of increased damming, which impairs freshwater-marine connectivity crucial for migratory and anadromous/catadromous species (e.g., eels, shads). Habitat fragmentation significantly weakens ecological resilience, particularly under warm conditions.
• The Environmental domain deteriorates substantially:

  ○

  Pollution (+0.12) and Temperature (+0.13) are strongly activated by climate change, highlighting heightened thermal stress, eutrophication, acidification, and pollutant accumulation.

  ○

  Nonindigenous Species also increase (+0.03), likely due to enhanced invisibility of ecosystems under altered temperatures and degraded habitats.

• Most biological traits worsen, reflecting lower resilience:

  ○

  Number of Offspring (−0.04), Reproductive Frequency (−0.05), Age at Maturity (−0.02), and Maximum Age (−0.04) decline, signaling an erosion of species’ reproductive capacities.

  ○

  These changes reflect ecosystem-wide pressures on survival, development, and recruitment.

  ○

  Only Mode of Reproduction shows a slight improvement (+0.03), which may reflect an adaptive shift toward more resilient strategies (e.g., internal fertilization), although this gain is negligible compared to broader biological stress.

• Social and technological factors (Overfishing, Poaching, Trawling, By-catch) remain nearly stable, with marginal negative changes (e.g., −0.01). This implies that even without intensification in these domains, their cumulative effect under the stress of climate and hydrological alteration can exacerbate ecological degradation.

This dual-pressure scenario emphasizes the synergistic amplification of extinction risk when climate drivers coincide with structural freshwater modifications like damming. While the change in the Extinction of Sea Life node is subtle, the underlying degradation across biological, environmental, and habitat dimensions is concerning. These findings advocate for integrated mitigation policies, targeting not only emissions reduction but also connectivity preservation in riverine and coastal systems. Failure to address both simultaneously may compromise long-term marine biodiversity in the Mediterranean.

4.4.4. Scenario: Intensification Climate Mitigation Success with Continued Damming (Climate Change = −1, Dams = +1)

Finally, this scenario examines a moderate management approach, where climate change is successfully mitigated (Climate Change = −1), but infrastructure expansion such as dam construction continues (Dams = +1). This configuration tests the hypothesis that controlling climate variables can substantially reverse biological and ecological degradation, even in the presence of structural stressors. As illustrated in Figure 11 we analyze the following results:

• Extinction of Sea Life shows a substantial decrease (−0.12), suggesting that mitigating climate change alone has a strong protective effect on marine biodiversity. This result is further supported by the overall improvement in biological and environmental variables.
• Living Region improves slightly (−0.06), indicating a partial buffer effect, even though dams continue to limit migratory pathways and ecosystem connectivity. The moderate impact suggests that while barriers remain ecologically disruptive, their effect is less pronounced in the absence of climate amplification.
• The Environmental domain experiences a notable reversal of degradation:

  ○

  Temperature (−0.31) and Pollution (−0.21) significantly decrease, both of which are highly responsive to reductions in climate pressures such as global warming and acidification.

  ○

  Nonindigenous Species remain nearly neutral (+0.03), likely due to ecosystem stabilization and lower stress, reducing invisibility.

• The Biological components exhibit strong positive responses:

  ○

  Reproductive Frequency (+0.16) and Number of Offspring (+0.13) increase markedly, suggesting enhanced reproductive potential and resilience.

  ○

  Maximum Age (+0.13) and Age at Maturity (+0.05) rise, indicative of population structures with longer lifespans and better recruitment conditions.

  ○

  Overall, life-history traits shift toward more stable and sustainable patterns under reduced environmental stress.

• Technological and social pressures remain largely unchanged, with Poaching, Trawling, Overfishing, and By-catch near-zero or marginally positive. This confirms that without direct regulation of these pressures, their impact persists, albeit less severe when ecosystem resilience is higher.

For this scenario, we conclude that it provides strong evidence that climate change mitigation has a profound systemic benefit, even in the continued presence of infrastructure-induced stress like damming. Key ecological indicators such as extinction risk, pollution, and reproductive health show meaningful recovery. However, dams continue to exert pressure, particularly on migratory species and freshwater-marine continuity. Policymakers are thus encouraged to view climate mitigation as foundational, but not sufficient alone—integrated management addressing both climate and structural interventions remains essential for long-term marine ecosystem sustainability.

FCMs have been widely applied in ecological modeling and environmental management [13,14,39], most studies rely either on expert knowledge or stakeholder perceptions without systematically integrating multidimensional socio-ecological drivers. The present work advances this field by embedding a modified PESTEL framework (BESTEL) and PCA within the FCM methodology. This integration offers a theoretical increment by reducing dimensionality, enhancing model robustness, and ensuring that both biological traits and socio-economic pressures are explicitly represented. To our knowledge, this is the first attempt to combine PCA-driven variable selection with an adapted PESTEL framework in an FCM-based decision-support system for marine biodiversity conservation, thereby bridging qualitative expertise and quantitative rigor in a single modeling platform.

A limitation of this study is that the expert panel, while comprising 34 specialists from diverse fields such as aquaculture, marine biology, ichthyology, and zoology, was largely concentrated in two Greek universities. This geographic and institutional concentration may introduce bias by underrepresenting perspectives from other Mediterranean regions or disciplines. To mitigate this, we ensured disciplinary diversity within the panel and complemented expert input with robust statistical methods (PCA) to reduce subjective weighting effects. Future research would benefit from expanding the geographic scope of experts to capture a broader range of ecological and policy contexts across the Mediterranean.

4.5. Expanded Analysis for Discussion

• Interpretation of Main Findings

The DSS results consistently highlight climate change and dam-related pressures as the two most influential drivers of marine species extinction in the Mediterranean. This outcome reflects the systemic and multi-dimensional impacts of these factors. Climate change exerts direct pressures on marine ecosystems through rising sea surface temperatures, ocean acidification, and altered current regimes, which disrupt reproductive cycles, migration patterns, and food availability. Similarly, dams fundamentally alter freshwater inflows, sediment deposition, and nutrient transport, creating long-lasting ecological disruptions in coastal and marine environments. By contrast, drivers such as economic development and legal frameworks, while important, were found to have a more indirect and secondary influence. This is because their effects are mediated through governance or market mechanisms rather than through immediate ecological disruption. The DSS simulations therefore confirm that biophysical and infrastructural pressures are the most critical levers in preventing further biodiversity loss.

• Comparisons with Literature

These findings are consistent with a growing body of literature emphasizing the disproportionate role of climate change in marine biodiversity decline. For instance, ref. [91] demonstrated that climate-related stressors account for over 50% of projected species loss in Mediterranean ecosystems. Our results extend these insights by integrating them within a structured DSS framework, confirming that these drivers remain dominant even when systematically evaluated against political, social, and economic dimensions. Interestingly, the relatively lower influence of economic and legal drivers differs from some earlier studies [92], which emphasized governance capacity as a central determinant of biodiversity outcomes. This divergence may be due to the expert panel’s emphasis on ecological interdependencies rather than institutional dynamics, a perspective that aligns more closely with the ecological literature than with policy-focused assessments.

• Policy Implications

The policy relevance of these results is substantial. First, they indicate that climate change adaptation and mitigation policies must remain a top priority for marine conservation in the Mediterranean. Strategies such as reducing greenhouse gas emissions, expanding marine protected areas with climate-resilient boundaries, and monitoring thermal refugia are critical interventions. Second, the DSS underscores the urgency of addressing dam-related pressures through integrated river basin management, rethinking dam siting, and where feasible, promoting dam removal or retrofitting to restore natural hydrological connectivity. Importantly, the systemic nature of the DSS allows decision-makers to recognize that interventions in these two domains can produce disproportionately high conservation benefits compared to isolated economic or social measures. The tool therefore enables prioritization, helping policymakers focus limited resources on the interventions most likely to produce measurable improvements in biodiversity outcomes.

• Sensitivity and Robustness Checks

To ensure that the DSS results are not artifacts of methodological design, sensitivity tests were conducted on the FCM weights and PCA component selection. Adjusting FCM weights by ±10% yielded similar dominance of climate change and dam-related factors, indicating the robustness of these findings. Likewise, varying the number of principal components retained in the PCA did not significantly alter the ranking of key drivers. These consistency checks strengthen confidence in the results and suggest that the DSS framework is stable under reasonable parameter variations. Future research could extend this robustness analysis by incorporating a larger and more geographically diverse expert panel, which would further validate the generalizability of the findings.

The consistency tests demonstrated that the DSS results were stable under reasonable parameter variations. Adjusting FCM weights by ±10% or altering PCA thresholds did not change the identification of climate change and dam-related pressures as the most dominant drivers. This robustness strengthens confidence in the findings and suggests that the DSS provides reliable support for policy prioritization in Mediterranean marine biodiversity management.

5. Conclusions

This study developed a comprehensive decision-support framework to understand and analyze the complex drivers behind the extinction of sea life in the Mediterranean. By combining Principal Component Analysis (PCA) and FCM, we offered a systemic approach that bridges empirical data, expert knowledge, and dynamic scenario simulation. The integration of these methods proved instrumental in exploring both direct and indirect causal relationships, facilitating the identification of key ecological vulnerabilities and policy leverage points.

The first stage of the research involved the application of PCA to reduce the dimensionality of biological and technological data. Through this statistical approach, we isolated a set of highly influential traits that are closely tied to species resilience and extinction vulnerability. Simultaneously, stressors like pollution, temperature increase, and human activities (e.g., trawling, damming) were highlighted as critical environmental and structural pressures that shape ecological outcomes. The insights from PCA not only guided concept selection but also informed the construction of a more parsimonious and focused FCM.

Building upon the PCA, the FCM was then constructed through a multi-phase participatory process. Concepts were drawn from both literature and expert surveys and organized within the BESTEL framework, encompassing biological, environmental, social, technological, economic, and legal dimensions. The model was refined through validation rounds with academic specialists in marine ecology and ichthyology, resulting in a reduced yet robust cognitive map of 17 interconnected concepts. These included key drivers like climate change and dams, along with outcome nodes such as pollution, temperature, living region, and extinction of sea life. The strength and direction of the relationships between concepts were quantified using de-fuzzified survey data, and the hyperbolic tangent activation function was employed to simulate the system’s dynamic behavior.

The true strength of the FCM emerged through the implementation of scenario analysis. We systematically manipulated the two main driver nodes (climate change and dams) to explore four plausible futures and their respective ecological consequences. In the most optimistic scenario, where both climate change and dam construction were minimized, the model projected a substantial reduction in the risk of marine species extinction. This outcome was not limited to the extinction concept alone, indeed there was a marked improvement that was observed across the system. Environmental concepts such as pollution and temperature declined sharply, suggesting that upstream interventions to restore ecological balance can yield cascading benefits. Additionally, biological traits like reproductive frequency and maximum age exhibited positive shifts, indicating enhanced reproductive potential and longevity for marine species. Even the “Living Region” concept, representing habitat accessibility and suitability, improved significantly, reinforcing the idea that structural and climatic restoration can revitalize marine ecosystems at both species and habitat levels. In contrast, a scenario where climate change remained unmitigated while dam impacts were reduced painted a more cautionary picture. Despite the removal of physical barriers and some improvement in habitat connectivity the extinction risk remained virtually unchanged. This was largely due to the dominant and far-reaching effects of climate change, which continued to exacerbate pollution, temperature, and the spread of nonindigenous species. Reproductive traits also declined under this condition, signaling that progress in one domain may be easily offset by deterioration in another, especially when it involves climate dynamics. The third scenario simulated the most critical trajectory under which a future where both climate change intensifies, and dam construction proliferates. Here, the cumulative pressures led to a modest but worrying increase in extinction probability. Temperature and pollution levels escalated, and biological indicators deteriorated across the board. Notably, habitat accessibility declined, illustrating the compounding effect of climate stress and habitat fragmentation. While some traits attempted to adapt, (for example there was a slight improvement in reproductive mode), these responses were marginal and insufficient to reverse the overall negative trend. Interestingly, the final scenario was characterized by successful climate mitigation but continued dam construction. This simulation yielded results that affirmed the centrality of climate as a systemic driver. Even with persistent structural pressures, the model indicated a substantial reduction in extinction risk, accompanied by improvements in reproductive traits, temperature, and pollution levels. While the “Living Region” showed only a moderate increase, the overall biological response was positive, suggesting that strategies which are focused on climate can significantly enhance ecosystem resilience, even in the face of ongoing infrastructure expansion.

Altogether, these scenario simulations revealed the differential and interactive impacts of environmental and structural pressures on marine biodiversity. The results underscored that while single domain interventions (such as dam removal) can yield targeted benefits, only integrated and upstream strategies that address climate change can produce broad and lasting improvements across ecological, biological, and habitat dimensions.

In summary, this study illustrates the value of combining data-driven methods like PCA with participatory modeling tools like FCM to understand complex ecological problems. The scenarios demonstrated that effective marine conservation in the Mediterranean requires a dual focus: climate mitigation as a foundational priority, and structural connectivity restoration as a critical complement.

Looking ahead, several avenues for future research emerge. First, the current FCM is static and operates under steady-state assumptions. Future efforts should integrate temporal dynamics, enabling the model to simulate ecological processes over time and observe how system responses evolve under different time horizons. Incorporating differential equations or agent-based modeling techniques could further enhance the model’s ability to capture nonlinear interactions, feedback loops, and time-lagged effects within the marine environment.

Second, future research could benefit from integrating spatial components into the FCM. By geo-referencing concepts such as “Pollution,” “Temperature,” and “Living Region,” researchers could simulate region-specific scenarios across different Mediterranean sub-basins (e.g., Aegean Sea, Adriatic Sea, or Levantine Basin), allowing for spatially explicit policy interventions. Coupling the FCM with Geographic Information Systems (GIS) or Earth Observation data (e.g., Sentinel or MODIS imagery) would enable real-time monitoring and improve the model’s environmental realism.

Third, the participatory process could be extended to include a broader range of stakeholders, such as fishers’ associations, NGOs, marine conservation agencies, and coastal communities. This would enhance the social legitimacy and policy relevance of the FCM, while also allowing for the inclusion of localized knowledge and value-driven priorities. Furthermore, conducting cross-national comparisons among Mediterranean countries would illuminate governance gaps and shared challenges, helping to inform a regional conservation framework grounded in system-based insights.