Systematic Review of Agent-Based and System Dynamics Models for Social-Ecological System Case Studies

: Social–ecological system (SES) modeling involves developing and/or applying models to investigate complex problems arising from the interactions between humans and natural systems. Among the different types, agent-based models (ABM) and system dynamics (SD) are prominent approaches in SES modeling. However, few SES models inﬂuence decision-making support and policymaking. The objectives of this study were to explore the application of ABM and SD in SES studies through a systematic review of published real-world case studies and determine the extent to which existing SES models inform policymaking processes. We identiﬁed 35 case studies using ABM, SD, or a hybrid of the two and found that each modeling approach shared commonalities that collectively contributed to the policymaking process, offering a comprehensive understanding of the intricate dynamics within SES, facilitating scenario exploration and policy testing, and fostering effective communication and stakeholder engagement. This study also suggests several improvements to chart a more effective trajectory for research in this ﬁeld


Introduction
The current challenges faced by society and the environment are both systemic and managerial [1].They are systemic because they stem from intricate and interconnected processes that operate at various scales, from local to global, and between various subsystems of the social and ecological realms.These issues cannot be fully comprehended through the lens of a single academic discipline.On the other hand, the issues are managerial because they require concerted and purposeful efforts by policymakers to address them in a sustained and coordinated manner.To address these challenges, a social-ecological systems (SES) approach has emerged that adopts a holistic systemic perspective towards the human and non-human elements [2].
SES refers to the complex and dynamic interrelationships between human societies and the natural environments within which they are embedded [3].The social component benefits from the services provided by an ecosystem and, in turn, human agency directly or indirectly modifies the functioning and structure of the ecosystem [4].The central question in SES research is concerned with understanding and managing the complex interactions between humans and nature across different scales and dimensions.This field emphasizes the interdependence and interconnectivity between social and ecological components, recognizing that human actions and decisions can have both positive and negative impacts on the environment and that these impacts can affect human well-being [5].SES research aims to enhance the resilience and sustainability of human-environment systems [6], which requires the integration of knowledge from different disciplines and perspectives and accounting for uncertainty and feedback [7].Several operational frameworks have been developed, such as the Panarchy framework, depicting system resilience as an outcome of connected adaptive cycles at different scales [8]; the conceptual cascade framework of "Pattern-Process-Service-Sustainability" that builds on understanding the coupled human and natural system [9]; a social-ecological framework for measuring the contributions of ecosystem services to society [10]; a diagnostic framework to assess the sustainable utilization and management of public resources [3]; a social-ecological action situation framework to analyze the emergence of social-ecological phenomena from system interactions [7]; an analytical framework of regime shifts in socialecological systems [11]; and three pillars of a sustainability framework that comprise social, economic, and environmental dimensions to provide a holistic perspective, ensuring that the complex interactions between human society and the natural environment are considered [12].SES frameworks have been used to comprehensively analyze key surface biophysical and socioeconomic processes and set the threshold of a safety boundary that generates more objective results [13].
Although frameworks define interactions and outcomes in SES, they are insufficient for facilitating scenario analysis, which is crucial for generating better social-environmental decisions and policies [14].Modeling techniques can be applied to simulate several scenarios.Computational models offer a systematic approach for conceptualizing real-world dynamics, the rational consequences of presumptions, past event patterns, and the outcomes of future situations [15].This may promote stakeholder buy-in by prompting evidence-based decision-making and shifting perceptions of the future to reflect realistic outcomes [16].Models are increasingly used to test the consequences of alternative assumptions about human behavior [17,18] or social-ecological relations [19] to elucidate the uncertainty associated with the complexity of human behavior and biophysical processes.Dynamic models have been widely used to study SES in various contexts and domains such as water resource management [20], fisheries [21], land use change [22], and urban development [23].Among the different types, the agent-based model (ABM) and system dynamics (SD) are two prominent approaches in complex system modeling [24].
ABMs simulate the behavior and interactions of individual agents in a system [25].These models have been used to explore, understand, explain, predict, communicate, illustrate, compare, and mediate social interactions among stakeholders or researchers from different disciplines [26].ABMs are often used to explore how individual-level behavior can impact the resilience or sustainability of a larger system [27].Regarding SES, ABMs are commonly employed for three primary purposes: (a) to explore and explain the emergence of social-ecological outcomes and understand how SES evolves over time; (b) to assess the impact of new policies or disturbances on a complex adaptive SES, encompassing potential unintended consequences; and (c) to facilitate participatory processes that enhance the comprehension of issues and collaborative problem-solving [28].In a SES, agents can represent individuals, households, organizations, and many more.Then, the model can simulate their decisions and interactions in response to environmental or social changes.ABMs have been used for SES modeling in irrigation systems [29], grazing systems [30], and coral reefs [31].
SD simulates the behavior of a system over time, focusing on feedback loops and interactions between different variables [32].SD modeling offers a set of conceptual, mathematical, and computational resources to address fundamental concepts in SES, such as feedback loops, nonlinear relationships, and regime shifts.SD modeling has been applied to explore the interconnections between system components, explicit representation of system-level dynamics through causal relationships, and responses of a SES to policy interventions and external forces [33].In a SES, SD can be used to explore the impacts of policy interventions or environmental changes on the overall system, identify leverage points for intervention, and for other applications [27].SD has been used to model SES in lake restoration [27], forest management [34], and coastal fisheries [35].
The value of SES modeling is largely determined by its applicability for understanding and interpreting real-world case studies [36].Case studies, which are widely used in SES research, can capture the diversity and complexity of SES by examining specific contexts that illustrate general patterns or principles [3].In addition, case studies can identify the key variables, indicators, drivers, outcomes, trade-offs, synergies, thresholds, and resilience of a SES and aid in developing sustainable policies and interventions [3,27].
SES modeling is emerging as a prominent research area, though it lacks appropriate research integration and synthesis [37].Previous reviews explored the modeling of the SES framework to identify the challenges [38], recommendations for good practice [39], strategies to advance reporting [40], and methodological guidelines for future applications [41].However, there has been no focused review on the integration of ABM or SD modeling outcomes in a particular case study related to the SES, specifically in the context of aiding the policymaking process.Consequently, only a few SES models have influenced decision support and policymaking compared to models from other areas such as transportation planning, epidemiology, and pesticide risk assessment [42][43][44].
This study aimed to explore the application of ABMs and SD in SES modeling through a systematic review of published real-world case studies and determine the extent to which existing SES models influence policymaking.We also explored the key characteristics of both modeling approaches for elucidating the complexity of the SES.Our comprehensive review elucidates the common factors associated with the improved integration of models into the policymaking process.

Materials and Methods
A systematic review of peer-reviewed literature was performed on 3 April 2023, using the scholarly databases Dimensions and Web of Science.We conducted a systematic review consistent with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [45] through four main steps (Figure 1).

Step 1: Systematic Literature Search in Dimensions and Web of Science
We searched for literature that modeled the SES framework using ABMs or SD in any real-world case study within the last 10 years.Although system modeling, both ABM and SD, and the idea of combining them are not new and date back to the late 1990s [46], we limited our article search only to the last 10 years to ensure that the review is an upto-date, comprehensive, and insightful analysis of the most recent research in the field.Both databases were searched using the following search terms in the title, abstract, and keywords fields: ("social-ecological system" AND "system dynamics" AND model*) for SD, and ("social-ecological system" AND "agent-based" AND model*) for ABMs.

Step 2: Screening of the Search Results
The list was refined by excluding duplicates, review articles, non-English articles, and articles without abstracts.The list was then manually refined by reading the abstracts and full text to check for applicability within the scope of this study.A study was considered eligible if it applied ABMs or SD to SES modeling within a case study.

Step 3: Coding of Included Publications for Data Collection
Each included article was read, evaluated, and coded using standardized criteria.In addition to the bibliometric information (i.e., author, title, year, and DOI), we collected complementary information by reviewing the abstract and full text of the final articles and coding them against four key aspects.

Step 1: Systematic Literature Search in Dimensions and Web of Science
We searched for literature that modeled the SES framework using ABMs or SD in any real-world case study within the last 10 years.Although system modeling, both ABM and SD, and the idea of combining them are not new and date back to the late 1990s [46], we limited our article search only to the last 10 years to ensure that the review is an up-todate, comprehensive, and insightful analysis of the most recent research in the field.Both databases were searched using the following search terms in the title, abstract, and keywords fields: ("social-ecological system" AND "system dynamics" AND model*) for SD, and ("social-ecological system" AND "agent-based" AND model*) for ABMs.

Aspect 1: Geographical Characteristics
Every SES requires a suitable system boundary to effectively identify management implications and facilitate comparisons [47].This is crucial because a SES should encompass the relevant ecological systems essential for maintaining key biophysical structures and processes.In doing so, the SES can provide ecosystem services and support social systems, including individuals and administrative bodies [48].To identify spatial boundaries, the articles were classified into four categories: local, regional, national, and international [49].In addition, the articles were classified based on their location, being in high-, upper middle-, lower middle-, or low-income countries, as previously described [50], to assess the difference in SES application between developing and developed countries.The characterization of a SES model aids in elucidating its intricate nature at a moderate level of abstraction between specific case studies and general theories, thereby facilitating a comparison and generalization of knowledge [51].In this study, the SES models were characterized based on a conceptual framework (Figure 2) structured into 13 dimensions distributed throughout the three main components of a SES: the social system, ecological system, and interactions between them [52].A list of variables for each dimension is presented in Table 1.This framework has two advantages.First, it is simple to understand by dividing the SES into three components that have several dimensions and variables.Second, this framework is quite general, making it suitable to describe the diverse SES model.
systems, including individuals and administrative bodies [48].To identify spatial boundaries, the articles were classified into four categories: local, regional, national, and international [49].In addition, the articles were classified based on their location, being in high-, upper middle-, lower middle-, or low-income countries, as previously described [50], to assess the difference in SES application between developing and developed countries.

Aspect 2: SES Component Being Modeled
The characterization of a SES model aids in elucidating its intricate nature at a moderate level of abstraction between specific case studies and general theories, thereby facilitating a comparison and generalization of knowledge [51].In this study, the SES models were characterized based on a conceptual framework (Figure 2) structured into 13 dimensions distributed throughout the three main components of a SES: the social system, ecological system, and interactions between them [52].A list of variables for each dimension is presented in Table 1.This framework has two advantages.First, it is simple to understand by dividing the SES into three components that have several dimensions and variables.Second, this framework is quite general, making it suitable to describe the diverse SES model.

Aspect 3: Stakeholder Involvement
The involvement of stakeholders could be underpinned in several ways, including through normative arguments (participation is a democratic right), substantive arguments (involvement produces better knowledge), instrumental arguments (participation improves the chance of success), and transformative arguments (improvement of social capital) [53].
Regardless, stakeholder involvement is critical for an impactful modeling endeavor [54].Given the increasing role of stakeholders in co-developing SES models, the type and extent of stakeholder participation in the reviewed studies were categorized as non-participatory, participation in model development, participation in model use, and participation in both model development and use.

Aspect 4: Practical Application from the Model
The practical application of SES models in the policymaking process can be evaluated based on their relevance to policy decision-making and legislative changes [15].If the model outcomes were relevant to policy decision-making or led to legislative changes, they represented a "high" level of practical application.In contrast, if the models primarily stimulate discussion and generate understanding without directly influencing policy decision-making, they were considered to have a "low" practical application [54].Furthermore, these models must have a user-friendly interface that effectively captures the complexity of the final models.This interface should be intuitive and easily navigable for end users to independently utilize the model [55].By ensuring user-friendliness, the model becomes more accessible for practical applications.

Step 4: Summary and Analysis of Collected Data
The identification of common factors would provide insights for future applications while also positing critical reflections on the limitations of current approaches.A comparative analysis of the reviewed literature was conducted to assess the potential applicability of the ABMs and SD of SES in the policymaking process.This involved summarizing and analyzing the four aspects across all articles using descriptive statistics and diagrams in Microsoft Excel (Microsoft 365 Apps for enterprise) and Quantyl Discovery (version 2.0).

Results
From the initial pool of 6242 papers, 81 were chosen for comprehensive full-text screening.Finally, 35 articles met the inclusion criteria, including 16 papers utilizing SD, 17 utilizing ABM, and two employing a hybrid SD-ABM approach.Appendix A provides the full details of the coding for the four key aspects.
Figure 3 shows the distribution of the study scales in relation to the income levels of the countries.Regional-scale studies have been conducted across all types of countries.Local-scale investigations were primarily concentrated in high-and upper-middle-income countries, whereas national-scale studies were more prevalent in upper-and lower-middleincome countries.
fewer studies conducted in upper-and lower-middle-income countries (23% and 17%, respectively).Only one study was conducted in a low-income country.
Figure 3 shows the distribution of the study scales in relation to the income levels of the countries.Regional-scale studies have been conducted across all types of countries.Local-scale investigations were primarily concentrated in high-and upper-middle-income countries, whereas national-scale studies were more prevalent in upper-and lower-middle-income countries.

Aspect 2. SES Component Being Modeled
Among the 13 dimensions encompassing the SES components, ABMs encompassed 10, albeit with varying degrees of emphasis (Figure 4a).Within the social component, particular attention was directed towards the well-being and development (WBD) dimension, whereas the governance (G) and human population dynamics (HPD) dimensions received limited consideration.Regarding the ecological component, considerable emphasis was placed on both organic carbon dynamics (OCD) and water dynamics (WD).In contrast, the nutrient cycling (NC) and disturbance regime (DR) dimensions were less prevalent, and the surface energy balance dimension was absent across the studies examined.SES interaction components exhibited nuanced patterns.Notably, ecosystem service supply (ESS) and human action on the environment (HAE) shared a substantial proportion of the interactions, signifying their interconnected nature.The ecosystem service demand (ESD) dimension occupied the remaining portion, whereas the ecosystem disservice supply (EDS) and socialecological coupling (SEC) dimensions were absent from the analyzed studies.Furthermore, the strong representation of the ESS and HAE dimensions facilitates a cohesive linkage between the social and ecological components.These dimensions effectively bridge the interaction between the human and ecological facets of the system, highlighting their intricate interdependence and underscoring the role of ABMs in capturing these vital connections within the SES.
SD incorporated 12 of the 13 intrinsic SES dimensions, albeit with varying degrees of emphasis (Figure 4b).Within the social component, the WBD and HPD dimensions were prominent and shared comparable representation, while the G dimension constituted the remaining portion of the social facet.Regarding the ecological component, the OCD dimension comprised 50%, the NC and WD dimensions shared 40%, while the DR dimension encompassed the remaining portion.Similar to the ABM findings, the surface energy balance dimension was absent.Within the interaction component, the SD prominently featured both the ESS and HAE dimensions, sharing nearly equivalent proportions.The ESD, EDS, and SEC dimensions collectively accounted for the remaining share.Notably, the ESS and HAE dimensions demonstrated a superior capacity for linking social and ecological The hybrid SD-ABM approach, applied in two articles, encompassed 7 of the 13 dimensions (Figure 4c).WBD and HPD predominated the social component; OCD, NC, and WD predominated the ecological component; and HAE and EDS predominated the interaction component.The hybrid SD-ABM approach effectively integrated these dimensions, contributing to a more comprehensive understanding of the intricate interplay between the social and ecological elements within the SES.
fectively bridge the interaction between the human and ecological facets of the system, highlighting their intricate interdependence and underscoring the role of ABMs in capturing these vital connections within the SES.
SD incorporated 12 of the 13 intrinsic SES dimensions, albeit with varying degrees of emphasis (Figure 4b).Within the social component, the WBD and HPD dimensions were prominent and shared comparable representation, while the G dimension constituted the remaining portion of the social facet.Regarding the ecological component, the OCD dimension comprised 50%, the NC and WD dimensions shared 40%, while the DR dimension encompassed the remaining portion.Similar to the ABM findings, the surface energy balance dimension was absent.Within the interaction component, the SD prominently featured both the ESS and HAE dimensions, sharing nearly equivalent proportions.The ESD, EDS, and SEC dimensions collectively accounted for the remaining share.Notably, the ESS and HAE dimensions demonstrated a superior capacity for linking social and ecological components, underscoring their pivotal role in enhancing cross-component interactions within the SES.The hybrid SD-ABM approach, applied in two articles, encompassed 7 of the 13 dimensions (Figure 4c).WBD and HPD predominated the social component; OCD, NC, and WD predominated the ecological component; and HAE and EDS predominated the interaction component.The hybrid SD-ABM approach effectively integrated these dimensions, contributing to a more comprehensive understanding of the intricate interplay between the social and ecological elements within the SES.

Aspect 3: Stakeholder Involvement
Table 2 shows the number of studies reviewed based on stakeholder involvement and modeling techniques.While stakeholder involvement is pivotal to the success of modeling endeavors [54], half of the reviewed articles (49%) did not incorporate stakeholders into their modeling processes.This omission can be attributed to the inherent technical intricacies of certain models, necessitating specialized knowledge that stakeholders may lack, potentially constraining the value of their participation [63,67,71,78,81].The importance of stakeholder involvement is often constrained in models that focus on ecological systems [64,72,73,79,83,85].Moreover, time and resource constraints, particularly for national-level studies [86,87] spanning jurisdictional boundaries [65] or requiring extensive data inputs [82], can render meaningful stakeholder engagement impractical.Interestingly, Bitterman and Bennett [76] noted that involving stakeholders could potentially enhance their models in future work.In the remaining articles, we observed notable variations in the extent of stakeholder engagement.Within this subset, 31% of the articles engaged stakeholders during model development to delineate problems and set boundaries, 11% incorporated stakeholders into the application phase to validate the model outcomes and ensure the robustness of the results through negotiation, and 9% embraced stakeholder involvement in both model development and application.
Analyzing the prevalence of stakeholder engagement in relation to system modeling techniques revealed pertinent insights.Among the reviewed articles employing SD, half (50%) did not integrate stakeholders into the modeling process.Conversely, the remaining 50% demonstrated diverse degrees of stakeholder involvement: 25% during model development, 19% during model application, and 6% during both the development and application phases.A parallel examination of articles employing ABMs revealed that 47% did not engage stakeholders, 41% involved stakeholders during model development, 6% during model application, and 6% during both phases.Of the two articles employing the hybrid SD-ABM approach, one abstained from stakeholder involvement, while the other embraced stakeholders during both model development and application.

Aspect 4: Practical Application from the Model
Table 3 lists the number of studies using practical applications and modeling techniques.The utility of SES models in the decision-making process was evident as 54% of the articles demonstrated a notable level of practical applicability.A defining attribute of system modeling is its capacity for scenario analysis through simulation, yielding quantifiable policy recommendations.This capability was effectively harnessed by a subset of the reviewed articles, with nine utilizing SD, eight employing ABMs, and two adopting a hybrid SD-ABM approach to highlight the advantageous outcomes achievable through scenario-based analyses.Within this cohort of 19 articles, only 5 incorporated a user-friendly interface, which has a considerable impact on model application by policymakers.Accessible interfaces empower policymakers and promote the effectiveness of the modeling endeavor, bolstering the role of SES models as a robust tool for evidence-based policy formulation and implementation and promoting informed engagement for harnessing the full analytical capabilities of SES models to meet the nuanced demands of real-world policy contexts.Consequently, the judicious inclusion of user-friendly interfaces amplifies the practical applicability of SES models and empowers policymakers to leverage their potential for shaping sustainable and effective policy outcomes.

Discussion
Our systematic review of research articles utilizing the SD, ABM, and hybrid SD-ABM approaches for SES modeling provides valuable insights into the potential contributions of these approaches to policymaking.By analyzing the breakdown of the models into social, ecological, and interaction components, we compared their strengths, weaknesses, and commonalities and elucidated the role of each modeling technique in enhancing our understanding of SES dynamics.

Strengths and Limitations of ABMs
Most ABMs with a high level of applicability focus extensively on the well-being and development dimensions within the social component.Its unique capacity to capture individual-level interactions and decision-making processes enables the simulation of agents' strategies for optimizing yields and income [57- 59,69,86].This granularity extends to the ecological component, where ABMs excel in depicting organic carbon and water dynamics.The model's proficiency in simulating intricate ecological processes, ranging from species movement to groundwater resource dynamicsprovides valuable insights [57, 59,61,62,64].Notably, ABMs effectively represent human actions in the environmental dimension, stemming from its inherent focus on emergent behaviors arising from agent interactions.This capacity aligns seamlessly with the modeling of human efforts in shaping their surroundings through activities such as land-use alterations, conservation initiatives, and restoration programs [57, 61,64,80,86].Furthermore, ABMs offer insights into the drivers, changing processes, and spatial characteristics within a SES, particularly through the simulation of individual agent interactions [30].In addition, ABMs can effectively inform managers about the trade-offs inherent in complex and diverse policy decisions by modeling individual heterogeneity, which is crucial for quantitatively evaluating the consequences of policies [65,74].
However, it is important to recognize that ABMs, while a powerful tool, should not serve as the sole determinant of policymaking, because complex social, political, and ecological aspects may not be adequately addressed by simulations alone [65].Moreover, the high computational demands of ABMs, particularly in complex or large-scale scenarios, can restrict its real-time application in policy analysis and decision-making processes [81].These demands, coupled with data limitations, can inadvertently hinder comprehensive stakeholder involvement, which is a key factor in successful policy integration [59].

Strengths and Limitations of SD
Conversely, SD exhibits a balanced representation across all SES components, with a notable emphasis on organic carbon dynamics, well-being and development, ecosystem service supply, and human actions on the environment.The ability of SD to capture feedback loops and dynamics renders it applicable for modeling nearly all SES dimensions, indicating its usefulness in elucidating the feedback mechanisms between social and ecological systems [87].Understanding these feedback mechanisms could inform the development of a holistic framework for SES management by facilitating the effective communication of scientific results to managers and guiding environmental decisionmaking through objective comparisons of different management options [75].Moreover, SD can integrate disparate data types over extended time periods, uncover robust connections between human and natural subsystems, and provide flexibility in exploring alternative scenarios [56].
The capacity to aggregate and average variables enhances the applicability of SD in modeling large-scale interactions and offers a high-level perspective on system behavior.However, this aggregation process can lead to oversimplification of complex interactions, potentially neglecting crucial system intricacies [82].It is important to acknowledge that, while versatile, SD should be applied with caution to avoid oversimplifying complex socio-ecological dynamics [88].SD also faces challenges in terms of time-series data availability [84].

Strengths and Limitations of the Hybrid SD-ABM
Hybrid models are emerging methodologies that combine the strengths of SD and ABMs, thereby facilitating the integration of macro-level dynamics with micro-level individual behaviors [26].Three methods exist for constructing a hybrid SD-ABM model: integrated, interfaced, and sequential hybrid designs [89].In integrated hybrid models, ABM and SD merge, allowing ABM and SD to interact simultaneously.Interfaced hybrids feature independent ABM and SD models exchanging data at designated simulation points.Sequential hybrids run ABM and SD separately, with one's output becoming the other's input.Both reviewed articles utilized the integrated hybrid design.Although the hybrid approach enhances the modeling of SES, the complexity introduced by the hybrid models can challenge stakeholder involvement because of the need to understand both modeling paradigms.As the demand for models that capture macro-and micro-level SES dynamics grows, the use of hybrid models is anticipated to increase.

Commonalities
Having their own strengths and limitations, these three modeling approaches share commonalities that collectively contribute to the policymaking process.

Comprehensive Insight and Integration
Our systematic review demonstrated that all three modeling approaches offer a comprehensive understanding of the intricate dynamics within SES.ABM could depict emergent behaviors arising from interactions among agents, offering valuable insights into the complex adaptive mechanisms within an SES.On the contrary, SD is better at grasping feedback loops and dynamics, demonstrating its effectiveness in unraveling the complex feedback mechanisms between social and ecological systems.Using these models, policymakers can gain a holistic view of how social and ecological components interact and influence each other.This integrated perspective provides a foundation for decisionmaking, allowing policymakers to recognize the complex relationships among human actions, ecological responses, and overall system behavior.The ability of these models to identify critical drivers and feedback loops further enriches this understanding, enabling policymakers to pinpoint areas in which interventions can be most effective and anticipate potential system responses.By capturing the interdependencies between social and ecological dimensions, these approaches emphasize the inseparable nature of human and natural dynamics in SES, urging policymakers to consider both facets simultaneously in policy formulation.

Policy Evaluation and Decision Support
One of the notable strengths shared by the SD, ABM, and hybrid SD-ABM is their capacity to facilitate scenario exploration and policy testing.These models allow policymakers to simulate a wide range of scenarios and policy interventions, offering a controlled environment for assessing potential outcomes, trade-offs, and unintended consequences.Furthermore, these approaches generate quantitative insights that enable objective datadriven decision-making.By incorporating empirical data and quantitative analysis, policymakers can formulate evidence-based strategies to increase the likelihood of achieving the desired policy outcomes.This quantitative approach also enables the assessment of trade-offs and synergies among various policy options, ensuring that policies are both effective and balanced in addressing the multiple dimensions of SES.

Effective Communication and Engagement
In addition to providing insights and decision-making support, the SD, ABM, and hybrid SD-ABM offer a suite of tools that foster effective communication and stakeholder engagement.Through visualization and scenario analysis, these models translate complex system dynamics into accessible visual representations, enabling policymakers to communicate trends, relationships, and potential policy impacts more effectively.This visualization aids in engaging stakeholders, including policymakers, communities, and interest groups, by offering a tangible platform for understanding the implications of different policy choices.Furthermore, the iterative nature of the models promotes adaptive management and learning.Policymakers can observe how a system responds to various interventions, thereby encouraging a dynamic and responsive approach to policy formulation.By involving stakeholders throughout the modeling process, from development to application, policymakers can ensure that decisions are informed by diverse perspectives, enhancing the legitimacy and acceptance of policies within the broader community.

Implications for Future Research
This review highlights the major achievements in the field of SES modeling in case studies employing SD and ABM.These modeling approaches have provided valuable insights into the complex dynamics of SES and their implications for policymaking.However, this analysis has certain limitations in current modeling paradigms.To advance the field and harness the full potential of ABM and SD for SES modeling, we provide several suggestions for future directions in this research field.
First, interdisciplinary collaboration among researchers should be fostered to improve data availability for model formalization.In this context, participatory modeling approaches would be valuable.Interdisciplinary teams and stakeholders can leverage diverse knowledge and perspectives to advance the capabilities of the models and ensure their relevance and applicability.Second, the integration of the ABM and SD approaches within hybrid models shows promise.The synergy between an ABM's micro-level focus on individual behaviors and SD's macro-level systemic insights can provide a more comprehensive understanding of SES dynamics.This hybridization can help overcome the oversimplification in SD models by capturing the finer details of interactions and behaviors.However, hybrid models can be more complex and less accessible to non-experts.To address these challenges, transparent model documentation and reporting practices should be developed to clearly outline how the SD and ABM components interact, which would enhance the credibility and replicability of the model.Finally, implementing machine learning (ML) algorithms in the ABM and SD can enhance their performance in modeling SES.In ABMs, ML algorithms can be used to develop more sophisticated agent behaviors and decision-making rules, whereas agents can learn from their interactions with the environment and other agents, allowing for the representation of adaptive and evolving behaviors [90].In SD, ML algorithms can be used to optimize model parameters, which can be particularly valuable in scenarios where finding the best parameter values is challenging [91].This optimization can promote the fit of empirical data and real-world observations to these models.Furthermore, ML algorithms can automate the generation and exploration of a wide range of scenarios in both ABMs and SD.This can help researchers and policymakers more efficiently assess the potential impacts of different policy interventions, management strategies, and environmental changes.

Limitations of This Study
While we acknowledge the crucial role of handling uncertainty in modeling, it should be noted that the inclusion of this aspect was not explicitly detailed in our review.We assumed that managing uncertainty is a standard practice in modeling, recognizing its significance in the robustness and reliability of model outcomes.However, it is imperative to recognize that addressing uncertainty can vary significantly based on the specific context, the nature of the uncertainty (such as parameter uncertainty or model structure uncertainty), and the objectives of the modeling exercise.Given the diverse and context-dependent nature of uncertainty-handling techniques, we chose to exclude this aspect from our review.This exclusion, therefore, represents a limitation of our study, highlighting the complexity and variability inherent in addressing uncertainty within the realm of SES modeling.Future research endeavors could delve into this critical dimension, exploring the nuanced techniques and methodologies employed in managing uncertainties to further enrich the understanding of modeling practices in the domain of SES.
Another limitation of our study is the depth of exploration into participatory modeling methodologies.Although our categorization encompassed various levels of stakeholder involvement in model development and utilization, we overlooked specific techniques such as mediated and companion modeling.Mediated modeling, where the modeler acts as a mediator between stakeholders and the model, and companion modeling, which emphasizes collaborative model construction, offer more profound insights through active stakeholder engagement.Thus, future research should delve deeper into these advanced participatory modeling approaches to offer a more comprehensive perspective on stakeholder involvement in modeling SES.

Figure 1 .
Figure 1.A PRISMA diagram of the identification and selection of studies.

Figure 1 .
Figure 1.A PRISMA diagram of the identification and selection of studies.

Figure 3 .
Figure 3. Study scale based on the level of income.

Figure 3 .
Figure 3. Study scale based on the level of income.
their pivotal role in enhancing cross-component interactions within the SES.

Figure 4 .
Figure 4. Alluvial diagrams showing the relationship between social and ecological systems by model type.(a) ABM, (b) SD, and (c) hybrid SD-ABM.Abbreviations: DR = disturbance regime; NC = nutrient cycling; WD = water dynamics; OCD = organic carbon dynamics; HPD = human population dynamics; G = governance; WBD = well-being and development; ESS = ecosystem service supply; ESD = ecosystem service demand; HAE = human action on the environment; EDS = ecosystem disservice supply; SEC = social-ecological coupling.

Figure 4 .
Figure 4. Alluvial diagrams showing the relationship between social and ecological systems by model type.(a) ABM, (b) SD, and (c) hybrid SD-ABM.Abbreviations: DR = disturbance regime; NC = nutrient cycling; WD = water dynamics; OCD = organic carbon dynamics; HPD = human population dynamics; G = governance; WBD = well-being and development; ESS = ecosystem service supply; ESD = ecosystem service demand; HAE = human action on the environment; EDS = ecosystem disservice supply; SEC = social-ecological coupling.

Table 2 .
Number of articles based on stakeholder involvement and modeling technique.

Table 3 .
Number of reviewed articles based on modeling technique and practical application.