Simulating endogenous institutional behaviour and policy implementation pathways within the land system Author links open overlay panel

Simulating endogenous institutional behaviour and policy implementation pathways within the land system Yongchao Zeng a,* , Joanna Raymond a , Calum Brown a,b , Mohamed Byari a , Mark Rounsevell a,c,d a Institute of Meteorology and Climate Research, Atmospheric Environmental Research (IMK-IFU), Karlsruhe Institute of Technology, 82467 Garmisch-Partenkirchen, Germany b Highlands Rewilding Limited, The Old School House, Bunloit, Drumnadrochit IV63 6XG, UK c Institute of Geography and Geo-ecology, Karlsruhe Institute of Technology, 76131 Karlsruhe, Germany d School of Geosciences, University of Edinburgh, Drummond Street, Edinburgh EH8 9XP, UK ARTICLE INFO Keywords: Land use change Ecosystem services Decision-making Agent-based modelling Institutional modelling Socio-ecosystem modelling ABSTRACT Policy interventions have substantial effects on land use change, providing key levers for multiple objectives, including mitigating climate change and biodiversity loss, and maintaining food security. Policy effects are often complicated, conflicting, and subject to regular change. Despite this, land system models typically treat policies as simple, exogenous modifications to models. To better represent the dynamic nature of policy-making, we develop an endogenous institutional model that can be embedded within land system models, here exemplified by an agent-based model. Numerical experiments are conducted to examine an institution with two policies targeting the production of ecosystem services. We find a clear scope for simulation-based exploration of policymaking, with emergent processes including the marginal diminishing effect of economic policy interventions, asymmetric spill-over effects for different ecosystem services, and trade-offs between policy goals. The endogenous institutional model demonstrates the potential to reveal various emergent patterns with important consequences for land systems. 1. Introduction Policy interventions in the land system must address a wide range of interacting processes to achieve ambitious but essential goals including climate change mitigation and adaptation (Guo et al., 2024), food security (Bengochea Paz et al., 2020) and biodiversity recovery (Broussard et al., 2023). Progress towards these goals has been fitful at best, and absent at worst, with many policies being counterproductive, mutually confounding, or subject to frequent changes that undermine their efficacy (Brown et al., 2019a; Lee et al., 2019). The EU’s Common Agricultural Policy alone provides a rich array of recent examples, with repeated attempts to balance food security and environmental protection largely failing, and the premature abandonment of controversial policies in the face of public opposition, such as farmer protests (Catalan News, 2024; European Court of Auditors, 2017). Land system models often aim to support policy-making by analysing the impact of public policy interventions on land use change (Berchoux et al., 2023; Li et al., 2017; Lippe et al., 2022). However, these models apply policy interventions exogenously and usually singly, and so are unable to account for the feedbacks that exist between policies, land users and policy institutions themselves (Lambin and Meyfroidt, 2010; Long and Qu, 2018). These models are thus inherently unable to capture fundamental characteristics of policy development and implementation. In particular, models neglect the dynamics of policy interventions and the causal relationships with land users that could be represented by endogenizing the policy process, and so remove important facets of realism including interactivity (Gonz´ alez, 2016; Ostrom, 2005), path dependency (Capoccia, 2015; Hegmon, 2017; Torfing, 2009), and complicatedness (Sun et al., 2016). Endogenous institutions have been increasingly recognised in the literature, particularly in areas such as common-pool resource (CPR) management, where institutions are often treated as sets of rules, norms, or strategies (Crawford and Ostrom, 1995). Research has demonstrated that institutions can emerge organically from the interactions of individual actors (Ostrom, 1999). This perspective has been especially useful for understanding how institutions form around resources such as land, water, and energy (Ghorbani et al., 2021, 2020). Modelling efforts have further enriched this view. For instance, Ghorbani et al. (2017) * Corresponding author. E-mail address: [email protected] (Y. Zeng). Contents lists available at ScienceDirect Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel https://doi.org/10.1016/j.ecolmodel.2025.111032 Received 29 July 2024; Received in revised form 22 December 2024; Accepted 17 January 2025 Ecological Modelling 501 (2025) 111032 Available online 23 January 2025 0304-3800/© 2025 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). developed the SONICOM model, an agent-based simulation that demonstrates how CPR users self-organise to develop sustainable rules through repeated interactions and adaptation, highlighting the importance of endogenous institutions. These emergent institutions play important roles in social-ecological systems, and their simulation highlights the scope and value of institutional modelling—particularly in demonstrating the long-term, systemwide impacts of interactions between modelled entities. Meanwhile, policies are often understood as official rules imposed by governing authorities. These official rules may synergise with or conflict with those arising organically from individual interactions, leading to varied policy intervention outcomes (Grzymala-Busse, 2010; Helmke and Levitsky, 2004). Building on the understanding of endogenous institutions as dynamic and interaction-driven, this research views institutions as policymakers interacting with the land system or broader socio-ecological systems. This perspective extends the focus from localised, emergent rules to institutions functioning as adaptive decision-makers at a systems level. Such institutions are endogenous in that they dynamically adapt to the land systems they operate within. Their policy actions are also influenced by the actions of numerous individual agents and, potentially, by emergent norms or rules developed from individual-level interactions. Nevertheless, examples of the modelling of institutions as decisionmakers within land-use systems are still quite limited (e.g., see Holzhauer et al.(2019)), which might be attributed to challenges such as the complexity of institutional decision-making processes, the potential for institutional change over time, the availability and accuracy of data, and issues of institutional transparency and accountability. In addition, computational complexity increases as the number of institutions and their interactions increase. Despite these difficulties, the rewards in terms of a more holistic representation of system dynamics make the endeavour worthwhile because it can unveil previously unexplored emergent phenomena and insights into institutional behaviour and its impact on land use change. As Davidson et al. (2024) highlighted, considering endogenous institutional changes in modelling is significant for understanding the socio-behavioural processes relevant to sustainability transitions. Several methodologies exist for simulating the decision-making behaviours of adaptive agents, among which machine learning and deep learning algorithms appear to be promising (Ramchandani et al., 2017). However, instead of using data-driven, black-box methods (Hu et al., 2023), we argue that modelling the complexities inherent in institutional decision-making should satisfy three guiding principles—parsimony, transparency, and extensibility (Loyola-Gonz´ alez, 2019; Sun et al., 2016) while recognising the central roles of heuristics and incrementalism in institutional decision-making processes (Gigerenzer et al., 2022; Pal, 2011). Heuristics is a vital component of both human and organisational decision-making processes (Gigerenzer et al., 2022, 2011). A significant feature of heuristic decision-making is its focus on limited information and the likely trade-off of optimality for speed (Gigerenzer and Gaissmaier, 2011; Gigerenzer and Goldstein, 1996; Kahneman, 2011; Russell and Norvig, 2010). Incrementalism is referred to as the science of “muddling through” (Lindblom, 1959), providing a model of policy-making through modest modifications rather than through comprehensive overhauls. Heuristics intrinsically resonates with incrementalism in political science on a theoretical basis (Dahl and Lindblom, 1965; Pal, 2011): both concepts illustrate the nature of bounded rationality in humans; moreover, incrementalism may represent a macroscopic manifestation of heuristic decision-making by policymakers. The objective of this research is to therefore create a modelling framework that endogenises the institutional decision-making process while respecting the above principles and the central roles of heuristics and incrementalism. This framework enables multiple institutions with different policy instruments and targets to respond to changes in the land system as it evolves in response to policies implemented, providing new insights into the interplay of institutional dynamics and land use changes. The institutional model is applied to the CRAFTY agent-based land use model (Murray-Rust et al., 2014) to examine the potential emergent patterns the institutional model can produce alongside behavioural land user agents. 2. Methods 2.1. Model overview Inspired by the work of Easton (1965) and Wlezien (1995), the role of endogenous institutions can be depicted as a sophisticated controller mechanism from a system perspective. Macroscopically, the outline structure of the endogenous institutional model coupled with a land use model (represented by the CRAFTY agent-based modelling framework) is a closed-loop control system, where an institution is populated with a sequence of components forming a decision entity. Within this system, institutions can observe and influence land use processes to achieve policy goals. Fig. 1 offers an overview of operational procedures within the loop. For illustrative clarity, these operational procedures are categorised into three parts, including the institutional model, the land use change model, and the part that solely represents the policy implementation procedure, the juncture where the two models tightly intersect. Within the institutional model, we adopted two further methodological approaches from control theory: Proportional-IntegralDerivative (PID) and fuzzy control (Carvajal et al., 2000; Kaur and Singh, 2019; Misir et al., 1996). A PID controller continually adjusts the disparity between a set point (e.g., a policy target) and the system’s existing state by factoring in three sources of error. In pursuit of policy goals, institutions can be modelled to adapt their decisions based on: 1) The current gap between the actual and desired policy outcomes (Proportional); 2) The accumulated impact of past policies and their resulting discrepancies (Integral); and 3) The changing speed with which these discrepancies are evolving (Derivative). Simulating policy adaptations based on these three types of gaps between the outcomes of interest and policy targets provides a simple yet systematic approach that mirrors the principle of heuristics and incrementalism in policymaking. A Fuzzy Logic Controller (FLC) serves as a function approximator that maps the goal-output discrepancies onto policy measures. The merits of coupling the PID and fuzzy control are manifold. An FLC is driven by an inference engine using a set of IF-THEN logic rules (Kaur and Singh, 2019). This rule-based paradigm fosters intuitive comprehension amongst human stakeholders and facilitates the encapsulation of knowledge from model users and policymakers. Compared with a sole PID controller, the joining of an FLC endows the modelled institutions with the capability of coping with nonlinear systems (Brown and Harris, 1995; Carvajal et al., 2000). Here, the PID controller allows for an adaptive, feedback-orientated approach to evaluating the disparity between an imposed policy goal and the model output, whilst the FLC maps the goal-output discrepancies onto policy adjustments. These approaches not only resonate with our core principles of institutional modelling but also offer practical algorithms that facilitate the effective operationalisation of endogenous institutional behaviours upon a solid theoretical basis. A policy implementation pathway comprises a series of timedependent policy actions, which might differ from “ideal policy actions” because of monetary and non-monetary constraints. For instance, IF an institutional agent intends to rapidly increase meat production, THEN it might decide to offer substantial subsidies. The “substantial subsidies” can be understood as an ideal policy action. However, due to the limitations of budgets and pressure from other stakeholders (arising, for example, from environmental concerns), the eventual subsidies for meat production may be far lower than intended, essentially resulting in a compromise policy action. In this case, the three factors - “ideal policy Y. Zeng et al. Ecological Modelling 501 (2025) 111032 2 actions”, budgetary constraints and stakeholder pressure – all contribute to the formation of policy actions and resultant policy implementation pathways, which tends to favour incremental policy adjustments. These mechanisms are detailed in Sections 2.3.2 and 2.3.3, where stakeholder pressure is generalised as policy inertia, which also reflects an institution’s resistance to change. 2.2. Model process The complete model process includes eleven operational steps as follows: 1. Initialise the land use model. 2. Initialise the institutional agents. This model can simulate many institutions simultaneously. Here, only one institution is shown for illustrative purposes. Within an institution, one crucial process is to initialise the policies, which includes defining the policy IDs, objectives, policy types, and any other policy characteristics to be included in the model. 3. Information collection by the institutions on whichever features of modelled land use are relevant (here, the demands and supplies of ecosystem services). Uncertainties might accompany the information collection. 4. Determine if it is time to adapt the policies. Policies remain unchanged within a set period of time that can represent, for example, election cycles. This step is considered here for three reasons. Firstly, institutions need time to allow the effects of the policies to become manifest and then to evaluate the outcomes. Secondly, institutions might have difficulties in responding to the changes with sufficient speed. Thirdly, policies might be designed in this way to gain more consistency. 5. If it is time to adapt the policies, the institution evaluates the performance of existing policies based on the collected information. The evaluation procedure uses the PID controller that considers the errors between the policy goals and actual outcomes. Optionally, the institution may also incorporate predicted errors into the evaluation. 6. Using the evaluation results, the institution conceives policy adaptations. A fuzzy logic module is applied to allow the integration of real-world policymakers’ knowledge into decisionmaking. The fuzzy logic module serves as a function that maps the evaluation results to policy adaptation. 7. A policy inertia constraint limits the magnitude of policy changes at each time step, which reflects the non-monetary (e. g., public opinion, interested parties, legislation) resistance to policy changes. 8. Subsequently, the institution deals with monetary constraints, i.e., budgets. In reality, institutional budgets can come from multiple sources and vary over time. The incorporation of a dynamic budget update process adds another layer of realism to the institutional model. 9. After updating the budget, the institution allocates the budget among different policies and outputs the formulated policy interventions. 10. The institution implements the policies in the land use system to push land use changes in the desired direction. 11. After the land use model processes the implemented policies, there is a check whether the end conditions are met. If true, then the simulation is stopped; otherwise, the information collected by the institutions is updated for the next iteration of decision-making. 2.3. Sub-models Further details about the various sub-models are provided here. To better present these details, the institutional model is segmented into four sub-models. The first sub-model focuses on the preliminary set-up and is limited to procedure 2: institution initialisation. The second sub-model, termed “information, evaluation, and adaptation”, encompasses procedures 3 to 7 due to their intrinsic link. This sub-model deals with information collection, uncertainty injection, policy evaluation, and adaptation together with the policy inertia constraint. The third submodel, termed “budget-allocation”, deals with updating the budget and allocating resources. The final sub-model focuses on policy implementation. Fig. 1. The operational procedures of the institutional model when embedded in a land use modelling framework. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 3 2.3.1. Sub-model 1: initialisation Each institution has a unique ID to distinguish itself from other institutions, a set that contains all policies available to this institution, a container to collect information, a list of variables to control the uncertainties, a set of variables and conditions defining its budget, and a set of decision rules (Fig. 2). A crucial task in this step is to initialise the policies and add them to the policy set. Each policy is essentially a group of attributes that can be adapted by the institution. The attributes/behaviours of institutions and policies as well as their relationships are shown in Fig. 2. The meanings of these behaviours and attributes are summarised in Tables A1 and A2 in Appendix A. The variables in the equations below are summarised in Table A3. Of crucial significance is the setting of unambiguous policy goals, as these lead to institutional adaptation throughout the simulation. Realworld examples of clearly stated policy goals can be found in the Paris Agreement (2015) regarding carbon emission reductions. Some specific examples include that, the United States and European Union have set goals to reduce greenhouse gas emissions by 2030 by 50–52 % compared to 2005 levels and by at least 55 % compared to 1990 levels, respectively (European Council, 2020; Zhao et al., 2022). These policies consistently specify a reference time, a deadline, and a targeted quantity. We use vector Gij =[Tij s,Tij e,Qij](1) to represent the goal of institution i’s policy j, which contains three components: Tij s the time when the policy starts, Tij e the time when the policy ends, and Qij the quantity a policy is meant to change from Tij s to Tij e. With the policy goals clearly defined, an appropriate initial policy intervention needs to be set up, which could be a real-world policy. For instance, if a simulation starts from the year 2020, the initial policy intervention could be the actual taxes and subsidies implemented in that year. The initial intervention can also be derived from model users’ intuitive estimation or deliberate calculation. 2.3.2. Sub-model 2: information, evaluation, and adaptation Institutions may have access to diverse sources of information to support decision-making processes, though the availability and quality of this information can vary depending on the context. While there are multiple sources of information available, gathering information can be resource-consuming, and the forms and extent of information can be limited as a result. Within the model, several categories of information are defined and represented as distinct data containers, labelled appropriately and filled with specific data points. Uncertainties can arise during information collection in reality, and so the collected data can be varied using defined value distributions, reflecting relevant forms of bias or error. Based on the information, institutions evaluate the state of the land use system relative to their goals. In reality, policies normally do not change frequently; it takes time for existing policies to manifest their impact and for institutions to respond to recent changes (Hocherman et al., 2024). Hence, a time lag is added to periodically trigger the evaluation and adaptation procedures for each policy (e.g. see Brown et al. (2019a) for a discussion of time lags in the land system). Time lags can be fixed or changed over time to reflect different triggering mechanisms of policy adaptation. A common example of the time lags in policy adaptations is election cycles. The evaluation of policy performance is a challenging task due to the complex nature of land use systems. It is difficult to attribute an outcome to a specific institutional action (Gonz´ alez, 2016). We adopt a heuristic approach to mimic institutional behaviour using a PID controller that adjusts the input based on the evaluation of three types of output-goal errors: proportional, integral and derivative. In this model, the proportional, integral, derivative errors, and their weighted sum are calculated using Eqs. (2), (3), (4), and (5) respectively: ε (P) tn=Qij −oij tn Qij (2) ε (I) tn=1 k∑ n m=n−k Qij −oij tm Qij (3) ε (D) tn=(Qij −oij tn)−(Qij −oij tn−k) kQij (4) E=C(P) ε (P) tn+C(I) ε (I) tn+C(D) ε (D) tn(5) where tn represents the specific time at which the institution evaluates the errors; oij tn is the output intended to be adjusted by institution i’s policy jat the time tn; k is the time interval of interest; ε (P) tn, ε (I) tn, Fig. 2. Institution and policy structures. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 4 ε (D) tnrespectively denote the proportional, integral, and derivative errors of policy j regarding its outcome oij tn at time tn. The weight vector [C(P) , C(I) ,C(D)], where C(P)+C(I)+C(D)=1 and C(P), C(I), C(D)∈ [0,1], can be applied to depict the policymakers’ sensitivity to these errors. The summation E of these errors, factoring in their respective weights compose the evaluation of the institution in terms of the performance of policy j. It is noteworthy that these errors can involve predicted outcomes, and thus, consider predicted errors in the calculation, depending on how institutions consider the reliability of the predictions. The FLC uses the weighted sum of errors E as an input, representing the performance evaluation of implemented policies. This controller works by mapping output errors onto policy adaptations, a crucial feature since institutions typically cannot directly influence the output but do so through policy instruments. As the FLC receives the input E, it is processed through three modules: fuzzification, inference engine, and defuzzification (Dadios, 2012). Fuzzification is the process that converts the crisp value of E into a set of fuzzy variables based on predefined membership functions. These fuzzy variables are defined over a range of values, allowing for a degree of membership rather than discrete categorisation. For instance, if the error E=0.1 representing the gap between the policy goal and actual crop production level is considered low here, this precise value might be classified under both the categories of ‘low’ and ‘high’ to extents determined by these membership functions. In this case, the fuzzification process might produce results such as “E = 0.1 belongs to ‘low’ with a membership degree of 0.9 and belongs to ‘high’ with a membership degree of 0.001 ″ , indicating E = 0.1 predominantly belongs to ‘low’. Subsequently, fuzzy inference maps the fuzzified E onto fuzzy output based on user-defined decision rules that are formatted in the IF-THEN structure. These rules are linguistic representations of experts’ knowledge, such as that of policymakers and researchers who have domain-specific interests. For example, a rule might be “IF E is low THEN the change of intervention of Policy j is small”. The third process is defuzzification, which translates the fuzzy output back to crisp real numbers again to allow the computer to process. The flexibility in adjusting membership functions, decision rules, and defuzzification algorithms allows fuzzy controllers to effectively capture and manipulate the relationships between various inputs and outputs in decision-making processes. Technically, the institutional agents’ behaviour in approaching policy goals is analogous to iterative approaches such as Newton’s method in solving ordinary differential equations (Cajori, 1911; Gal´ antai, 2000; Ypma, 1995), but a critical difference is that the institutions do not know the precise mathematical representation of the target system and hence need to conduct a series of constrained trial and errors to approach the policy goals. Also, it should be noted that the FLC is used to map the errors onto the incremental quantity of policy adjustments rather than onto a direct value indicating the intensity of the policy adjustments, reflecting the approach of incrementalism. Let F denote the function of FLC and F(E)indicate the incremental, quantitative adjustment to the existing policy, such as the changes in taxes, subsidies, geographical expansion of new protected areas. The policy adjustment is constrained by the policy inertia constraint Nij, a variable whose value is prescribed to reflect the non-monetary resistance to policy adjustments. The constrained policy adjustment at t +1 is denoted as Aij t+1and calculated using Equation (6). The sign function outputs the sign of its input. Aij t+1 is accumulated to form a policy modifier denoted as Mij t+1, as shown in Equation (7). It might be convenient to use normalised policy adjustment together with a fixed step size for iterative policy adaptation. In this way, the policy modifier is a coefficient of the step size. As shown in Equation (8), ƞij is the step size, and Vij t+1 is the modified policy adjustment for the (t+1)-th iteration. Aij t+1=sign(F(E)) × min(|F(E)|,Nij)(6) Mij t+1=Mij t+Aij t+1(7) Vij t+1= η ij ×Mij t+1(8) 2.3.3. Sub-model 3: budget allocation In modelling an institution with multiple policies, it is crucial to understand how much budget each has access to because the distribution of budget among institutions or policies is related to the power they can leverage to impact land use change or even other institutions. Hence, a process that updates the budget for an institution has been included in the model. The budget update process tracks the institution’s income and expenditure whenever a policy is applied. The institution can allocate the budget across multiple policies. It is assumed here that the intensity of a policy intervention is quantitively measurable, and that its absolute value is positively correlated to the budget the institution uses to implement the policy. As seen in Eq. (9), f is a monotone function that maps the absolute value of a policy intervention Vij t+1 to the resource Rij t+1 consumed. For simplicity, in the simulation section below, function f can be approximated as a linear function, and only subsidies are considered budget-consuming. f(Vij t+1)=Rij t+1(9) The allocation of the budget can be treated as an optimisation problem in quadratic form, which is a convex optimisation problem: min r∈(rmin,rmax)∑ j ξij(rij t+1−Rij t+1)2(10) s.t.0≤∑ j rij t+1≤Bi t(11) ∑ j ξij =1 (12) where Rij t+1 denotes the resource needed by institution i’s to implement policy j; rij t is the decision variable determining the resource allocated to implement policy j; ξij is a weight reflecting the comparative importance of policy j perceived by institution i; Bi t is the total budget of the institution i at t. The optimiser is intended to find a combination of rij t that minimise the objective function. Alternatively, one might consider using IF-THEN rules instead of optimisation to determine the resource allocation, or modifying the weights to add more dynamics. While IF-THEN rules are a feasible approach to budget allocation here, framing the budget allocation as a convex optimisation problem offers several distinct advantages. First, it provides a unique optimal solution, ensuring manageability and clarity even with numerous policies involved. Second, this method standardises the allocation process, reducing the need for extensive parameterisation compared to IF-THEN rules. Third, it allows model users to focus on customizing the decision rules for policy adaptation. Additionally, this approach simplifies the interpretation of results, as the adaptability of institutional agents is primarily determined by the policy adjustment rules, rather than two sets of rules in different processes. After finding the optimal resource allocation, it has to be transformed back to the policy intervention using the inverse function of f, as shown in Eq. (13): V∗ij t+1=sign(Vij t+1)f−1(r∗ij t+1)(13) where r∗ij t+1 is the optimal resource for Vij t+1; the sign function returns the sign of Vij t+1;V∗ij t+1 is the resultant optimal policy intervention. Because the policies consume the budget, the budget Bi t should be updated accordingly using Eq. (14): Y. Zeng et al. Ecological Modelling 501 (2025) 111032 5 Bi t←Bi t−∑ j r∗ij t+1(14) 2.3.4. Sub-model 4: policy implementation The policy implementation sub-model is an intersection of the institutional model and the land use change model. This part is highly customisable and should be coupled with the specifics of the land use change model. Here we use the CRAFTY model (Murray-Rust et al., 2014) for this purpose. In CRAFTY, land managers of various Agent Functional Types (AFTs) manage unique combinations of productivity across multiple ecosystem services. These managers utilise the resources within their land to produce ecosystem services and compete to meet the societal demand for their respective services. There is a diversity of policy instruments that can influence land use changes, among which economic measures play a crucial role. In this paper, we focus on the intervention of economic policies within the land use system represented by CRAFTY. Typically, economic policies include taxes and subsidies. In realworld cases, when economic policies come into play, the equilibrium of demand and supply is determined by both the elasticities of the demand and supply, even if the policies are imposed on one side of the market. For instance, subsidies on the supply side may cause a price drop, which in turn induces more demand and causes friction on the price drop. Because the current land use change model is focused on the supply side of different land use types and uses prescribed demands for different ecosystem services, it is assumed that taxes and subsidies are only imposed on the land users rather than the ecosystem service consumers. That is, the demand is assumed to be completely inelastic. The economic policies are implemented as follows: cxy =∑ S(pS(∑ i ViS t+1+mS)) (15) where cxy denotes the competitiveness of a land use agent at the land cell whose coordinates are (x,y); S is the ecosystem service the land user produces. pS is the land user agent’s production level of ecosystem service S within the land cell. ViS t+1 is the institution i’s economic policy that targets ecosystem service S; mS is marginal utility brought by ecosystem service S. Transitions between land uses, and hence changes in the supply levels of different services, are constrained by parameters representing opportunity costs and other barriers, here set at a single low level to prevent unrealistically high rates of change. 2.4. Experimental settings Although the institutional model allows for the incorporation of many institutions, it is not within this paper’s scope to demonstrate the model’s descriptive strength in modelling a variety of combinations of institutions and their policies. Instead, the experiments here serve as a proof of concept to explore what meaningful patterns can emerge at the system level and the interpretability of the model’s micro mechanisms, given the significance of emergent patterns in socio-ecological system modelling (Grimm et al., 2005; Jakoby et al., 2014; Kramer-Schadt et al., 2007; Piou et al., 2009). Hence, the parameterisation of the institutional model is set to be parsimonious to facilitate the understanding, interpretation, and evaluation of the model. Specifically, the numerical experiments here focus on the dynamics of one institution possessing two policies. Additional institutions and policies can be added in the same way. The experimental institution can be regarded as an agricultural policymaker responsible for direct interventions that address agricultural ecosystem services. The institution can affect two policies - Policy 1.1 and Policy 1.2 - that influence the production of meat and crops using taxes and subsidies. The simulated institution can use three types of decision rules labelled Economic, Tax, and Subsidy. In principle, decision rules can be arbitrarily complex. The current decision rules are designed to be a set of single-input-singleoutput functions that are straightforward enough for intuitive understanding. The detailed parametrisation of these decision rules and numerical settings are given in Tables B1–B3 in Appendix B using FLC language defined in the IEC 61131-7 (PLCopen, 2000). Moreover, the institution is assumed to evaluate only the integral errors to mitigate the influence of noise in the model. The institutional model is applied to a newly developed CRAFTY emulator implemented in Java (Zeng, 2024c). The emulator utilises the MASON agent-based modelling framework (Luke et al., 2019) and enables rapid adaptation for exploratory research. The fuzzy logic controllers are implemented using the jFuzzyLogic library (Cingolani and Alcala-Fdez, 2012; Cingolani and Alcal´ a-Fdez, 2013). The emulator is built based on the CRAFTY-EU land use model (Brown et al., 2019b) and parametrised with data based on the climatic and socio-economic scenario (Brown et al., 2019b). These data define the change in demands over time for each of the ecosystem services. Each AFT has a matrix of sensitivities to 8 land capitals and represents a type of land manager that can provide a range of services: meat, crops, diversity, timber, carbon, urban development and recreation. AFTs also have a sequence of production levels corresponding to the ecosystem services, which function similarly to the total factor productivity in a Cobb-Douglas production function. The names of the AFTs are shown in Table C1 in Appendix C. The initial distribution of the AFTs and the AFT attributes are given in Figs. C1 and C2. 3. Results 3.1. Baseline simulation To have a baseline understanding of how the land use model behaves, the model was run 100 times without institutional influence. Fig. 3 shows the model’s output in terms of the supply of different ecosystem services over time. It can be seen that the model’s output is stable across the simulations; except for timber production, the supply of all other ecosystem services shows a significant tendency to follow the demands, which means the model’s macroscopic adaptive behaviours emerge from the competition between different AFTs. It should be noted that there are only 71 years (from the 0th to 70th step) of scenario-based data to update the annual demands and capitals, after which these are held static until 150 steps have been completed (i.e. in total covering the 0th to 149th) to manifest the model’s equilibrium states. In the following experiments, the 70th and 149th years are respectively labelled as t1 and t2 for illustrative convenience. Timber production is an exception in this scenario because the timber yield capital decreases rapidly. That is, the deviation of timber production from the demand is caused by the limitation of relevant capital instead of by the model’s inherent mechanism. In sum, the natural behaviour of the model is featured by the tendency to close the gap between supply and demand, which offers a clear-cut baseline scenario to understand the influence of institutions. 3.2. The individual impact of Policy 1.1 and Policy 1.2 We examine the influence of the institution’s policies individually. The purpose of this experiment is to investigate the effectiveness of the institutional agents in functioning as actors that influence land use change as expected. To concentrate on this purpose, other influences are isolated: the budget constraints are temporarily deactivated and the fuzzy decision rules are set to “Economic”. This means the policies are not restricted to subsidies or taxes but are purely negative rectifiers aimed at closing the gap between policy goals and actual supply. The parametrisation details of the institution, Policy 1.1, and Policy 1.2 are given in Table 1. To probe the land use system’s reaction to different policy goals, a sequence of policy goals for the supply of specific services ranging from Y. Zeng et al. Ecological Modelling 501 (2025) 111032 6 0 to 6 with equal intervals of 0.1 are examined, resulting in 61 different policy goals. The results of applying a meat tax and subsidizing crop production are shown in Figs. 4 and 5, respectively. Combining the colours indicating different goals and the cluster of supply curves in each sub-figure, an evident trend can be observed: higher policy goals lead to higher production, and vice versa. This trend demonstrates that demand curves are no longer the only forces influencing the ecosystem service supply. In addition to this intuitive trend, there are three patterns worth noting: 1) different long-term and short-term impacts, 2) marginal diminishing effect, and 3) asymmetric spill-over effects. The difference between long-term and short-term policy effects can be observed by comparing the vertical width of the cluster of supply curves at t1 and t2. At t1, the lowest supply of meat reaches approximately 1.5 times the initial supply, and the highest point is around 3.5 times. Although the supply notably deviates from the original demand curve, gaps still exist between the supply and policy goals. At t2, the lowest and highest supply reaches approximately 1 and 4.5 times the initial supply, respectively, signifying that the policy interventions are still in effect and leading the system closer to the policy goals. The same trend can also be observed in Fig. 5 for crop production. These patterns result from the fact that the institution applies an incremental adaptation strategy and that the land use system takes time to respond to policy interventions. Fig. 6 presents the marginal diminishing effect more clearly. The actual supply achieved is concentrated around the demands. As the policy goals deviate from the demands along the horizontal axes, the discrepancies between the demands and actual supply become large, which signifies that the policy goals indeed influence ecosystem service Fig. 3. Supply and demand of different ecosystem services without policy interventions. The 70th and 149th year are labelled as t1 and t2, respectively, marking the end of the scenario-based changes in input data and the subsequent period of static input data respectively. Table 1 Parameterisation of the institution, Policy 1.1, and Policy 1.2. Experimental variables are highlighted in bold. Institution parameter Value Unique ID 1 Policies 1.1, 1.2 Information Crop supply and demand, meat supply and demand. Uncertainties Null Budget Unlimited Decision rules Economic Policy parameter First policy Second policy Unique ID 1.1 1.2 Target service Meat Crops Policy Type Economic Economic Initial guess 1,000,000 1,000,000 Policy inertia constraint 0.2 0.2 Policy goal 0.0 – 6.0 0.0 – 6.0 Intervention 0.0 0.0 Intervention modifier 0.0 0.0 Evaluation result 0.0 0.0 Time lag 5 5 Timer Equal to Time lag Equal to Time lag Adapting False False Y. Zeng et al. Ecological Modelling 501 (2025) 111032 7 production. However, as the policy goals move farther from the demands, the gap between the expected and actual supplies becomes larger, forming a flattened S shape that reflects the marginal diminishing effect of the institutional interventions. The results imply that the demand is a strong attraction for the supply to follow and can exert significant force against the policy interventions if the demand-goal discrepancies are significant. From Figs. 4 and 5, a cross-service impact can be observed and labelled as an asymmetric spill-over effect. While the policies imposed on meat production have an insignificant spill-over effect on the other types of ecosystem services, the policies on crop production have a nonnegligible impact on the meat supply. The asymmetric effect of policy interventions implies distinct AFT transitioning processes occurring “under the hood”. As most AFTs can produce multiple ecosystem services (at different levels), it is intriguing that the system can find a way to respond to the policy interventions on meat production while maintaining other ecosystem services including crop production almost at initial levels. Contrastingly, the system cannot react similarly to the crop policies, which causes a considerable spill-over effect on meat production. Comparing the gaps in the two sub-figures of Fig. 6, it can be seen that it is more challenging to achieve the policy targets of Policy 1.2 than Policy 1.1. The asymmetric spill-over effect captured in the above experiments indicates differences in the underlying AFT dynamics. Fig. 7 displays six dominant AFTs across different policy goals. The numbers of different AFTs were recorded at the end of each simulation. It can be seen that the multifunctional (Multifun) AFT plays a major role in both Policy 1.1 and Policy 1.2 experiments, while its number does not show a substantial overall change. This is plausible because the multifunctional AFT is a major contributor to carbon sequestration (see Fig. C2), for which the demand is almost constant in the scenario. In Policy 1.1, the numbers of intensive arable (IA) and mixed farming (Mix_Fa) agents experience notable decreases as the policy goal rises. In contrast, the intensive farming (Int_Fa) AFT, as a productive meat supplier (see Fig. C2), starts from a low quantity but becomes the most dominant type when the policy goal for meat production is highest. In terms of Policy 1.2, as the goal of crop production increases, the numbers of intensive pastoral (IP) and extensive agro-forestry (Ext_AF) agents decrease, while the multifunctional, intensive arable, and intensive farming AFTs increase (Fig. 7). Nevertheless, intensive pastoral and extensive agro-forestry agents are not crop producers; the other three AFTs can produce considerable quantities of crops and meat, which makes the meat supply deviate from the demand. Fig. 4. The impact of Policy 1.1, in which the institution applies an economic policy to meat production with a policy goal ranging from 0 to 6 times the initial meat supply. The red line represents the demand. The 70th and 149th year are labelled as t1 and t2, respectively. Another pattern is the marginal diminishing effect of the policy interventions, which can be seen from two features of the cluster of supply curves. The first feature is that the lower and upper bounds of the supply curves exhibit a diminishing speed to approach the lowest and highest policy goals. The second feature is the uneven distribution of the supply curves. The supply curves tend to approach the lower and upper bounds, which leaves the space in between less dense. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 8 Fig. 5. The impact of Policy 1.2, in which the institution provides an economic incentive for crop production with a policy goal ranging from 0 to 6 times the initial crop supply. The red line represents the demand. The 70th and 149th year are labelled as t1 and t2, respectively. Fig. 6. Relative supply under different policy goals at t1 and t2. The horizontal axes represent the goals of Policy 1.1 adjusting meat production and Policy 1.2 adjusting crop production, while the vertical axes are the ratio of the actual supply to the initial supply. The black and green solid lines, respectively, indicate the supply of the corresponding ecosystem services at t1 and t2. The dashed straight lines show the expected supply ratio. The vertical solid lines indicate the final demands at t1. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 9 Table A4 Parameterisation of the institution, Policy 1.1, and Policy 1.2. Experimental variables are highlighted in bold. Institution parameter Value Unique ID 1 Policies 1.1, 1.2 Information Crop supply and demand, meat supply and demand. Uncertainties Null Budget Gain from taxation Decision rules Tax or Subsidy Policy parameter First policy Second policy Unique ID 1.1 1.2 Target service Meat Crops Policy Type Tax Subsidy Initial guess 100,000 100,000 Policy inertia constraint 0.2 0.2 Policy goal 1.0 4.0 Intervention 0.0 0.0 Intervention modifier 0.0 0.0 Evaluation result 0.0 0.0 Time lag 1 - 10 1 - 10 Timer Equal to Time lag Equal to Time lag Adapting False False Appendix B Table B1 Parameterisation of decision rule labelled as “Economic”. FUNCTION_BLOCK economic VAR_INPUT gap: REAL; END_VAR VAR_OUTPUT Intervention: REAL; END_VAR FUZZIFY gap TERM nhigh:=(−0.5,1) (−0.3,0); TERM nmild:=(−0.5,0) (−0.3,1) (−0.1,0); TERM nlight:=(−0.3,0) (−0.1,1) (0,0); TERM neutral:=(−0.05,0) (0,1) (0.05,0); TERM plight:=(0, 0) (0.1, 1) (0.3,0); TERM pmild:=(0.1,0) (0.3,1) (0.5,0); TERM phigh:=(0.3, 0) (0.5, 1); END_FUZZIFY DEFUZZIFY intervention TERM nhigh:=(−0.2,1) (−0.1,0); TERM nmild:=(−0.15,0) (−0.05,1) (0,0); TERM neutral:=(−0.02,0) (0,1) (0.02,0); TERM pmild:=(0,0) (0.05,1) (0.15,0); TERM phigh:=(0.1,0) (0.2,1); METHOD: COG; DEFAULT:=0; END_DEFUZZIFY RULEBLOCK No1 AND: MIN; ACT: MIN; ACCU: MAX; RULE 1: IF gap IS nhigh THEN intervention IS nhigh; RULE 2: IF gap IS nmild THEN intervention IS nmild; RULE 3: IF gap IS nlight THEN intervention IS neutral; RULE 4: IF gap IS neutral THEN intervention IS neutral; RULE 5: IF gap IS plight THEN intervention IS neutral; RULE 6: IF gap IS pmild THEN intervention IS pmild; RULE 7: IF gap IS phigh THEN intervention IS phigh; END_RULEBLOCK END_FUNCTION_BLOCK Y. Zeng et al. Ecological Modelling 501 (2025) 111032 16 Table B2 Parameterisation of decision rule labelled as “Subsidy”. FUNCTION_BLOCK Subsidy VAR_INPUT gap: REAL; END_VAR VAR_OUTPUT intervention: REAL; END_VAR FUZZIFY gap TERM plow:=(0,1) (0.15,0); TERM plight:=(0.025, 0) (0.175, 1) (0.325,0); TERM pmild:=(0.175,0) (0.325,1) (0.45,0); TERM phigh:=(0.325, 0) (0.45, 1); END_FUZZIFY DEFUZZIFY intervention TERM neutral:=(−0.015,1) (0.05,0); TERM plight:=(0.025,0) (0.075,1) (0.125,0); TERM pmild:=(0.075,0) (0.125,1) (0.175,0); TERM phigh:=(0.125,0) (0.2,1); METHOD: COG; DEFAULT:=0; END_DEFUZZIFY RULEBLOCK No1 AND: MIN; ACT: MIN; ACCU: MAX; RULE 0: IF gap IS plow THEN intervention IS neutral; RULE 1: IF gap IS plight THEN intervention IS plight; RULE 2: IF gap IS pmild THEN intervention IS pmild; RULE 3: IF gap IS phigh THEN intervention IS phigh; END_RULEBLOCK END_FUNCTION_BLOCK Table B3 Parameterisation of decision rule labelled as “Tax”. FUNCTION_BLOCK Tax VAR_INPUT gap: REAL; END_VAR VAR_OUTPUT intervention: REAL; END_VAR FUZZIFY gap TERM nhigh:=(−0.5,1) (−0.3,0); TERM nmild:=(−0.4,0) (−0.3,1) (−0.2,0); TERM nlight:=(−0.3,0) (−0.1,1) (0,0); TERM nlow:=(−0.1,0) (0,1); END_FUZZIFY DEFUZZIFY intervention TERM neutral:=(−0.05,0) (0,1) (0.025,1); TERM light:=(−0.125,0) (−0.075,1) (−0.025,0); TERM mild:=(−0.175,0)(−0.125,1) (−0.075,0); TERM high:=(−0.2,1) (−0.125,0); METHOD: COG; DEFAULT:=0; END_DEFUZZIFY RULEBLOCK No1 AND: MIN; ACT: MIN; ACCU: MAX; RULE 1: IF gap IS nhigh THEN intervention IS high; RULE 2: IF gap IS nmild THEN intervention IS mild; RULE 3: IF gap IS nlight THEN intervention IS light; RULE 4: IF gap IS nlow THEN intervention IS neutral; END_RULEBLOCK END_FUNCTION_BLOCK Y. Zeng et al. Ecological Modelling 501 (2025) 111032 17 Appendix C Table C1 Names of the AFTs. Abbreviation Full Name VEP Very extensive pastoral Int_AF Intensive agro-forestry Mix_P Mixed pastoral IP Intensive pastoral Min_man Minimal management EP Extensive pastoral Ext_AF Extensive agro-forestry UMF Unmanaged forest Multifun Multifunctional Mix_For Mixed forest UL Unmanaged land IA Intensive arable MF Managed forest Mix_Fa Mixed farming Int_Fa Intensive farming Ur Urban P-Ur Peri-urban Fig. C1. Initial distribution of AFTs in CRAFTY-EU. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 18 Fig. C2. Normalised ecosystem service production levels of the 17 AFTs in CRAFTY-EU. Fig. C3. Transitions of all AFTs under different policy goals at the end of the model simulation at t2. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 19 Table C2 Potential approaches to parameterizing the institutional model in future research. Parameter/ Potential data source Stakeholder involvement Historical data Narrative-based simulation Empirical modelling data Sensitivity analysis Institution Policies ✔ ✔ ✔   Information ✔✔ ✔  Uncertainties ✔✔✔ Budget ✔ ✔ ✔ ✔ ✔ Decision rules ✔✔  Policy evaluation ✔✔✔ Budget allocation ✔✔✔ Policy Target service ✔ ✔ ✔   Policy type ✔ ✔ ✔   Initial guess ✔ ✔ ✔ ✔ Policy inertia ✔✔ ✔ ✔ Policy goal ✔ ✔ ✔ ✔ ✔ Time lag ✔ ✔ ✔ ✔ ✔ Appendix D Example of Parameterizing the Institutional Model: Potential Application to Organic Farming in Germany The example provided here is intended to enhance understanding of parameterisation and to illustrate guidance for the empirical application of this model. It might not be sufficiently detailed for a direct application, as this would require substantially more calibration and verification. Policy Goal To parameterise the institutional model, policy goals can be taken directly from published policy documents. In this example, the Organic Action Plan targets 25 % of EU agricultural land to be under organic farming by 2030 (European Commission, 2024). For Germany, the national target is more ambitious, with organic farming expected to cover 30 % of the total agricultural area by 2030. (For unquantified policy goals, the framework proposed by Fetting (2020), involving stakeholder input, could be used to establish measurable targets.) Information Historical data can be used to initialise key parameters. For example, the share of organically farmed land in Germany increased from 1.6 % in 1994 to 11.4 % by the end of 2023 (Kuhnert, 2024). This historical trend provides a basis for estimating the gap between the target (30 %) and the current share of organic farming. Spatial information on crop management (at least at state resolution) can be used to establish baseline distributions that affect subsequent uptake of organic farming (Kuhnert, 2024). Uncertainties may occur during real-world information collection, so the collected data can be adjusted and tested using predefined value distributions to account for relevant biases or errors. Budget According to Becker et al. (2022) and Lampkin and Sanders (2022), in Germany, organic farming is supported under the Second Pillar of the CAP Strategic Plans (2023–2027), which allocates approximately 20 % of its funding to organic farming. The second pillar represents about 10 % of Germany’s total CAP budget of over € 30 billion for the period. Based on these figures, the approximate annual funding for organic farming in Germany can be inferred. This budget can be distributed evenly across the years or adjusted to follow an expected growth trajectory for organic farming areas. Time Lag The institutional model should include time lags to account for the delayed effects of policy changes. Historical CAP funding periods, such as 2014–2020 (European Council, 2019) and 2023–2027 (European Council, 2024), suggest a time lag of 5–7 years for major adjustments. However, the EU Strategy on Adaptation to Climate Change (Climate ADAPT, 2024) advocates for faster responses, indicating that shorter time lags (e. g., 2–3 years) might also be meaningful. Initial Guess When historical data is unavailable for a new policy, an initial parameter estimate can be inferred. For organic farming, conversion subsidies for arable land in Germany average € 394/ha/year and maintenance subsidies are € 264/ha/year (Kuhnert, 2023). These values can serve as initial subsidy levels for the model. Policy Inertia Historical data on subsidies for organic farming can be analysed to determine the maximum and minimum annual changes. This range can be used to approximate policy inertia, which can help constrain the model to realistic variations and avoid erratic behaviour. Sensitivity analyses can further test the impact of different levels of policy inertia on model outputs. Decision Rules, Policy Evaluation, and Budget Allocation These elements involve subjective judgment by policymakers. Hypothetical settings can be tested to align with intuitive decision-making. For instance, a straightforward decision rule could be: IF the gap between the policy goal and the current outcome is large, THEN adjust the policy significantly. For policy evaluation, historical trends in organic farming coverage, which show steady growth with minimal fluctuations, suggest that using the latest coverage of organic farming land might be enough to estimate the gap between the policy goal and actual coverage. Budget allocation can be estimated based on historical funding allocation within the CAP framework if multiple policies sharing the same budget account are considered. For greater accuracy, in-depth interviews with stakeholders—such as policymakers from the Federal Ministry of Food and Agriculture or experts from the Thünen Institute (¨ Ozdemir, 2022)—could further clarify strategies for policy adjustments and budget planning. Data availability I have shared the links to my data and code in the manuscript. References Agreement, P., 2015. The Paris agreement, The report of the Conference of the Parties to the United Nations Framework Convention On Climate Change. HeinOnline, pp. 1–60. Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C.S., Jansson, B.-O., Levin, S., M¨ aler, K.-G., Perrings, C., Pimentel, D., 1995. Economic growth, carrying capacity, and the environment. Ecolog. Econ. 15, 91–95. https://doi.org/10.1016/ 0921-8009(95)00059-3. Balcan, M.-F., Sharma, D., 2024. Learning Accurate and Interpretable Decision Trees. arXiv preprint. https://doi.org/10.48550/arXiv.2405.15911. Batini, N., Nelson, E., 2001. The lag from monetary policy actions to inflation: friedman revisited. Int. Finance. 4, 381–400. https://doi.org/10.1111/1468-2362.00079. Becker, S., Grajewski, R., Rehburg, P., 2022. Where does the CAP money go? Design and priorities of the draft CAP Strategic Plans 2023–2027.https://doi.org/10.3 220/WP1655118238000. Bekaert, G., Hoerova, M., Lo Duca, M., 2013. Risk, uncertainty and monetary policy. J. Monet. Econ. 60, 771–788. https://doi.org/10.1016/j.jmoneco.2013.06.003. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 20 Bengochea Paz, D., Henderson, K., Loreau, M., 2020. Agricultural land use and the sustainability of social-ecological systems. Ecol. Modell. 437, 109312. https://doi. org/10.1016/j.ecolmodel.2020.109312. Benson, D., Russel, D., 2015. Patterns of EU energy policy outputs: incrementalism or punctuated equilibrium? West. Eur. Polit. 38, 185–205. https://doi.org/10.1080/ 01402382.2014.936707. Berchoux, T., Hutton, C.W., Hensengerth, O., Voepel, H.E., Tri, V.P.D., 2023. Effect of planning policies on land use dynamics and livelihood opportunities under global environmental change: evidence from the Mekong Delta. Land Use Policy 131, 106752. https://doi.org/10.1016/j.landusepol.2023.106752. Box, G.E.P., 1976. Science and statistics. J. Am. Stat. Assoc. 71, 791–799. https://doi. org/10.1080/01621459.1976.10480949. Bravo-Pe˜ na, F., Yoder, L., 2024. Agrobiodiversity and smallholder resilience: a scoping review. J. Environ. Manage. 351, 119882. https://doi.org/10.1016/j. jenvman.2023.119882. Broussard, A., Dahdouh-Guebas, F., Hug´ e, J., 2023. Diversity of perspectives in biodiversity conservation: a case study of port land use in Antwerp and Rotterdam. J. Environ. Manage. 341, 117937. https://doi.org/10.1016/j.jenvman.2023.117937. Brown, C., Alexander, P., Arneth, A., Holman, I., Rounsevell, M., 2019a. Achievement of Paris climate goals unlikely due to time lags in the land system. Nat. Clim. Chang. 9, 203–208. https://doi.org/10.1038/s41558-019-0400-5. Brown, C., Seo, B., Rounsevell, M., 2019b. Societal breakdown as an emergent property of large-scale behavioural models of land use change. Earth Syst. Dyn. 10, 809–845. https://doi.org/10.5194/esd-10-809-2019. Brown, M., Harris, C., 1995. Neurofuzzy Adaptive Modelling and Control. Prentice Hall International, UK) Ltd. Buchan, R., Cloutier, D.S., Friedman, A., 2019. Transformative incrementalism: planning for transformative change in local food systems. Prog. Plann. 134, 100424. https:// doi.org/10.1016/j.progress.2018.07.002. Cajori, F., 1911. Historical note on the Newton-Raphson method of approximation. Am. Math. Month. 18, 29–32. https://doi.org/10.1080/00029890.1911.11997596. Capoccia, G., 2015. Critical Junctures and Institutional change, Advances in Comparative-Historical Analysis. Cambridge University Press, pp. 147–179. Carvajal, J., Chen, G.R., Ogmen, H., 2000. Fuzzy PID controller: design, performance evaluation, and stability analysis. Inf. Sci. (Ny) 123, 249–270. https://doi.org/ 10.1016/S0020-0255(99)00127-9. Catalan News, 2024. EU Agricultural Policy Under Fire as Protesting Farmers Shake Green Pledges. Cingolani, P., Alcala-Fdez, J., 2012. jFuzzyLogic: a robust and flexible Fuzzy-Logic inference system language implementation. In: 2012 IEEE International Conference on Fuzzy Systems. IEEE, pp. 1–8. Cingolani, P., Alcal´ a-Fdez, J., 2013. jFuzzyLogic: a java library to design fuzzy logic controllers according to the standard for fuzzy control programming. Int. J. Comput. Intell. Syst. 6, 61–75. https://doi.org/10.1080/18756891.2013.818190. Clark, J., 2006. The institutional limits to multifunctional agriculture: subnational governance and regional systems of innovation. Environ. Plann. C: Govern. Policy 24, 331–349. https://doi.org/10.1068/c053. Climate ADAPT, 2024. A New EU Adaptation Strategy. https://climate-adapt.eea.europa. eu/en/eu-adaptation-policy/strategy. Convention on Biological Diversity, 2020. Aichi Biodiversity Targets. https://www.cbd. int/sp/targets. Crawford, S.E.S., Ostrom, E., 1995. A Grammar of Institutions. Am. Politi. Sci. Rev. 89, 582–600. https://doi.org/10.2307/2082975. Czy˙ zewski, B., Matuszczak, A., Grzelak, A., Guth, M., Majchrzak, A., 2021. Environmental sustainable value in agriculture revisited: how does Common Agricultural Policy contribute to eco-efficiency? Sustain. Sci. 16, 137–152. https:// doi.org/10.1007/s11625-020-00834-6. D’Acunto, F., Hoang, D., Paloviita, M., Weber, M., 2021. Effective Policy Communication: Targets Versus Instruments. https://www.econstor.eu/handle/104 19/232543. Dadios, E., 2012. Fuzzy Logic-Controls, Concepts, Theories and Applications. IntechOpen. https://doi.org/10.5772/2662. Dahl, R.A., Lindblom, C.E., 1965. Politics, Economics and welfare; Planning and PoliticoEconomic Systems Resolved Into Basic Social Processes. Harper & Row. Davidson, M.R., Filatova, T., Peng, W., Verbeek, L., Kucuksayacigil, F., 2024. Simulating institutional heterogeneity in sustainability science. Proceed. Natl. Acad. Sci. 121, e2215674121. https://doi.org/10.1073/pnas.2215674121. Del Monte-Luna, P., Brook, B.W., Zetina-Rej´ on, M.J., Cruz-Escalona, V.H., 2004. The carrying capacity of ecosystems. Glob. Ecol. Biogeogr. 13, 485–495. https://doi.org/ 10.1111/j.1466-822X.2004.00131.x. Easton, D., 1965. A systems analysis of political life. Etheredge, L.S., Short, J., 1983. Thinking about government learning. J. Manag. Stud. 20. https://doi.org/10.1111/j.1467-6486.1983.tb00197.x. European Commission, 2024. Agriculture and Rural Development. Action Plan for Organic Production in the EU. https://agriculture.ec.europa.eu/farming/organic-f arming/organic-action-plan_en. European Council, 2019. Reform of the Common Agricultural Policy post 2013. https: //www.consilium.europa.eu/en/policies/cap-reform/. European Council, 2020. Climate Change: What the EU is Doing. https://www.consilium. europa.eu/en/policies/climate-change/#2030. European Council, 2024. Common Agricultural Policy 2023–2027. https://www.consiliu m.europa.eu/en/policies/cap-introduction/cap-future-2020-common-agriculturalpolicy-2023-2027/. European Court of Auditors, 2017. Greening: a More Complex Income Support Scheme, not yet Environmentally Effective. https://www.eca.europa.eu/lists/ecadocuments/ sr17_21/sr_greening_en.pdf. Fetting, C., 2020. The European Green Deal. https://www.esdn.eu/fileadmin/ESDN_R eports/ESDN_Report_2_2020.pdf. Forssell, D., Powers, W., 2009. Perceptual control theory. Living Contr. Syst. Pub. 72, 74–75. Funk, C., Raghavan Sathyan, A., Winker, P., Breuer, L., 2020. Changing climate - changing livelihood: smallholder’s perceptions and adaption strategies. J. Environ. Manage. 259, 109702. https://doi.org/10.1016/j.jenvman.2019.109702. Gal´ antai, A., 2000. The theory of Newton’s method. J. Comput. Appl. Math. 124, 25–44. https://doi.org/10.1016/S0377-0427(00)00435-0. Ghorbani, A., Bravo, G., Frey, U., Theesfeld, I., 2017. Self-organization in the commons: an empirically-tested model. Environ. Modell. Softw. 96, 30–45. https://doi.org/ 10.1016/j.envsoft.2017.06.039. Ghorbani, A., Ho, P., Bravo, G., 2021. Institutional form versus function in a common property context: the credibility thesis tested through an agent-based model. Land Use Policy 102, 105237. https://doi.org/10.1016/j.landusepol.2020.105237. Ghorbani, A., Nascimento, L., Filatova, T., 2020. Growing community energy initiatives from the bottom up: simulating the role of behavioural attitudes and leadership in the Netherlands. Energy Res. Soc. Sci. 70, 101782. https://doi.org/10.1016/j. erss.2020.101782. Ghorbani, A., Siddiki, S., Bravo, G., 2023. Institutional adaptation and transformation for climate resilience. Front. Environ. Sci. 11, 1159923. https://doi.org/10.3389/ fenvs.2023.1159923. Gigerenzer, G., Gaissmaier, W., 2011. Heuristic decision making. Annu. Rev. Psychol. 62, 451–482. https://doi.org/10.1146/annurev-psych-120709-145346. Gigerenzer, G., Goldstein, D.G., 1996. Reasoning the fast and frugal way: models of bounded rationality. Psychol. Rev. 103, 650–669. https://doi.org/10.1037/0033295X.103.4.650. Gigerenzer, G., Reb, J., Luan, S., 2022. Smart heuristics for individuals, teams, and organizations. Ann. Rev. Organ. Psychol. Organ. Behav. 9, 171–198. https://doi.org/ 10.1146/annurev-orgpsych-012420-090506. Gigerenzer, G.E., Hertwig, R.E., Pachur, T.E., 2011. Heuristics: The foundations of Adaptive Behavior. Oxford University Press. https://doi.org/10.1093/acprof:oso/ 9780199744282.001.0001. Givel, M., 2010. The evolution of the theoretical foundations of punctuated equilibrium theory in public policy. Rev. Policy Res. 27, 187–198. https://doi.org/10.1111/ j.1541-1338.2009.00437.x. Gonz´ alez, V.B., 2016. Modelling Adaptation Strategies for Swedish Forestry Under Climate and Global Change. University of Edinburgh. Gonz´ alez, V.B., 2017. Modelling Adaptation Strategies for Swedish Forestry under Climate And Global Change. http://hdl.handle.net/1842/25380. Grimm, V., Revilla, E., Berger, U., Jeltsch, F., Mooij, W.M., Railsback, S.F., Thulke, H.-H., Weiner, J., Wiegand, T., DeAngelis, D.L., 2005. Pattern-oriented modeling of agentbased complex systems: lessons from ecology. Science. 310, 987–991. https://doi. org/10.1126/science.1116681. Grzymala-Busse, A., 2010. The best laid plans: the impact of informal rules on formal institutions in transitional regimes. Stud. Comp. Int. Dev. 45, 311–333. https://doi. org/10.1007/s12116-010-9071-y. Guidara, A., 2020. Policy Decision Modeling With Fuzzy Logic: Theoretical and Computational Aspects. Springer Nature. Guo, J., Li, F.Y., Tuvshintogtokh, I., Niu, J., Li, H., Shen, B., Wang, Y., 2024. Past dynamics and future prediction of the impacts of land use cover change and climate change on landscape ecological risk across the Mongolian plateau. J. Environ. Manage. 355, 120365. https://doi.org/10.1016/j.jenvman.2024.120365. Haug, N., Dan, S., Mergel, I., 2024. Digitally-induced change in the public sector: a systematic review and research agenda. Public Manag. Rev. 26, 1963–1987. https:// doi.org/10.1080/14719037.2023.2234917. Heald, D.A., 2002. Public policy towards cross subsidy. Ann. Public Cooperat. Econ. 68, 591–623. https://doi.org/10.1111/1467-8292.00066. Hegmon, M., 2017. Path Dependence The Oxford Handbook of Southwest Archaeology, pp. 1–13. https://doi.org/10.1093/oxfordhb/9780199978427.013.7. Helmke, G., Levitsky, S., 2004. Informal institutions and comparative politics: a research agenda. Perspect. Polit. 2, 725–740. https://doi.org/10.1017/S1537592704040472. Hocherman, T., Trop, T., Ghermandi, A., 2024. Time lags in environmental governance – a critical review. SSRN. https://doi.org/10.2139/ssrn.4689647. Holzhauer, S., Brown, C., Rounsevell, M., 2019. Modelling dynamic effects of multi-scale institutions on land use change. Reg. Environ. Change. 19, 733–746. https://doi.org/ 10.1007/s10113-018-1424-5. Howlett, M., Capano, G., Ramesh, M., 2018. Designing for robustness: surprise, agility and improvisation in policy design. Policy Soc. 37, 405–421. https://doi.org/ 10.1080/14494035.2018.1504488. Hu, T.X., Zhang, X.S., Bohrer, G., Liu, Y.L., Zhou, Y.Y., Martin, J., Li, Y., Zhao, K.G., 2023. Crop yield prediction via explainable AI and interpretable machine learning: dangers of black box models for evaluating climate change impacts on crop yield. Agric. For. Meteorol. 336, 109458. https://doi.org/10.1016/j.agrformet.2023.109458. Huang, S.-L., Lee, Y.-C., Chiang, L.-Y., 2022. Assessing the synergies and trade-offs of development projects in response to climate change in an urban region. J. Environ. Manage. 319, 115731. https://doi.org/10.1016/j.jenvman.2022.115731. Jakoby, O., Grimm, V., Frank, K., 2014. Pattern-oriented parameterization of general models for ecological application: towards realistic evaluations of management approaches. Ecol. Modell. 275, 78–88. https://doi.org/10.1016/j. ecolmodel.2013.12.009. Kahneman, D., 2011. Thinking, Fast and slow: Farrar. Farrar. Straus and Giroux. Kasprzyk, J.R., Nataraj, S., Reed, P.M., Lempert, R.J., 2013. Many objective robust decision making for complex environmental systems undergoing change. Environ. Modell. Softw. 42, 55–71. https://doi.org/10.1016/j.envsoft.2012.12.007. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 21 Kaur, R., Singh, A., 2019. Fuzzy logic: an overview of different application areas. Adv. Applic. Math. Sci. 18, 677–689. Knickel, K., Kr¨ oger, M., Bruckmeier, K., Engwall, Y., 2009. The challenge of evaluating policies for promoting the multifunctionality of agriculture: when ‘Good’ questions cannot be addressed quantitatively and ‘quantitative answers are not that good’. J. Environ. Policy Plann. 11, 347–367. https://doi.org/10.1080/ 15239080903033945. Kowalewski, S.A., Birch, J., 2020. How do people get big things done? In: Bondarenko, D. M., Kowalewski, S.A., Small, D.B. (Eds.), The Evolution of Social Institutions: Interdisciplinary Perspectives. Springer International Publishing, Cham, pp. 29–50. Kramer-Schadt, S., Revilla, E., Wiegand, T., Grimm, V., 2007. Patterns for parameters in simulation models. Ecol. Modell. 204, 553–556. https://doi.org/10.1016/j. ecolmodel.2007.01.018. Kratzke, W.P., 2009. Tax subsidies, third-party-payments, and cross-subsidization: America’s distorted health care markets. U. Mem. L. Rev. 40, 279–422. Kuhnert, H., 2023. Area-Based Support Payments for Organic Farming. https://www. thuenen.de/en/thuenen-topics/organic-farming/aktuelle-trends-der-deutschen-oek obranche/flaechenbezogene-foerderung-des-oekologischen-landbaus. Kuhnert, H., 2024. Organic farming in figures. https://www.thuenen.de/en/thuenen-to pics/organic-farming/aktuelle-trends-der-deutschen-oekobranche/oekolandbau-inzahlen. Kulovesi, K., Oberthür, S., 2020. Assessing the EU’s 2030 climate and energy policy framework: incremental change toward radical transformation? Review of European. Comparat. Int. Environ. Law 29, 151–166. https://doi.org/10.1111/ reel.12358. Lambin, E.F., Meyfroidt, P., 2010. Land use transitions: socio-ecological feedback versus socio-economic change. Land Use Policy 27, 108–118. https://doi.org/10.1016/j. landusepol.2009.09.003. Lampkin, N., Sanders, J., 2022. Policy Support for Organic Farming in the European Union 2010-2020. https://doi.org/10.3220/WP1663067402000. Lans Bovenberg, A., Goulder, L.H., 2002. Chapter 23 - Environmental Taxation and Regulation*. In: Auerbach, A.J., Feldstein, M. (Eds.), Handbook of Public Economics. Elsevier, pp. 1471–1545. Lee, H., Brown, C., Seo, B., Holman, I., Audsley, E., Cojocaru, G., Rounsevell, M., 2019. Implementing land-based mitigation to achieve the Paris Agreement in Europe requires food system transformation. Environ. Res. Lett. 14, 104009. https://doi. org/10.1088/1748-9326/ab3744. Li, J., Rodriguez, D., Tang, X.Y., 2017. Effects of land lease policy on changes in land use, mechanization and agricultural pollution. Land Use Policy 64, 405–413. https://doi. org/10.1016/j.landusepol.2017.03.008. Lindblom, C.E., 1959. The science of "Muddling through". Public. Adm. Rev. 19, 79–88. https://doi.org/10.2307/973677. Lippe, M., Rummel, L., Günter, S., 2022. Simulating land use and land cover change under contrasting levels of policy enforcement and its spatially-explicit impact on tropical forest landscapes in Ecuador. Land Use Policy 119, 106207. https://doi.org/ 10.1016/j.landusepol.2022.106207. Long, H., Qu, Y., 2018. Land use transitions and land management: a mutual feedback perspective. Land Use Policy 74, 111–120. https://doi.org/10.1016/j. landusepol.2017.03.021. Loyola-Gonz´ alez, O., 2019. Black-box vs. white-box: understanding their advantages and weaknesses from a practical point of view. IEEE Access 7, 154096–154113. https:// doi.org/10.1109/ACCESS.2019.2949286. Luke, S., Simon, R., Crooks, A., Wang, H., Wei, E., Freelan, D., Spagnuolo, C., Scarano, V., Cordasco, G., Cioffi-Revilla, C., 2019. The MASON simulation toolkit: past, present, and future. In: Davidsson, P., Verhagen, H. (Eds.), Multi-Agent-Based Simulation XIX. Springer International Publishing, Cham, pp. 75–86. Makate, C., Makate, M., Mango, N., Siziba, S., 2019. Increasing resilience of smallholder farmers to climate change through multiple adoption of proven climate-smart agriculture innovations. Lessons from Southern Africa. J. Environ. Manage. 231, 858–868. https://doi.org/10.1016/j.jenvman.2018.10.069. Marsden, G., Docherty, I., 2021. Mega-disruptions and policy change: lessons from the mobility sector in response to the Covid-19 pandemic in the UK. Transp. Policy. (Oxf.) 110, 86–97. https://doi.org/10.1016/j.tranpol.2021.05.015. Miller, J.H., Page, S.E., 2009. Complex Adaptive systems: an Introduction to Computational Models of Social life: an Introduction to Computational Models of Social Life. Princeton university press. https://doi.org/10.1515/9781400835522. Misir, D., Malki, H.A., Chen, G.R., 1996. Design and analysis of a fuzzy proportionalintegral-derivative controller. Fuzzy Set. Syst. 79, 297–314. https://doi.org/ 10.1016/0165-0114(95)00149-2. Mondal, C., Giri, B.C., 2023. Tax-subsidy or reward-penalty? Determining optimal strategy in sustainable closed-loop supply chain under quality-dependent return. Int. J. Syst. Sci.: Operat. Logist. 10, 2116738. https://doi.org/10.1080/ 23302674.2022.2116738. Moyson, S., Scholten, P., Weible, C.M., 2017. Policy learning and policy change: theorizing their relations from different perspectives. Policy Soc. 36, 161–177. https://doi.org/10.1080/14494035.2017.1331879. Murray-Rust, D., Brown, C., van Vliet, J., Alam, S.J., Robinson, D.T., Verburg, P.H., Rounsevell, M., 2014. Combining agent functional types, capitals and services to model land use dynamics. Environ. Modell. Softw. 59, 187–201. https://doi.org/ 10.1016/j.envsoft.2014.05.019. Osoba, O.A., Vardavas, R., Grana, J., Zutshi, R., Jaycocks, A., 2020. Modeling agent behaviors for policy analysis via reinforcement learning. In: 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA). IEEE, pp. 213–219. Ostrom, E., 1999. Coping with tragedies of the commons. Ann. Rev. Politi. Sci. 2, 493–535. https://doi.org/10.1146/annurev.polisci.2.1.493. Ostrom, E., 2005. Understanding Institutional Diversity. Princeton university press. https://doi.org/10.2307/j.ctt7s7wm. ¨ Ozdemir, C., 2022. Organic Farming in Germany. https://www.ovgu.de/unimagdeburg _media/Forschung+_+Wirtschaft/Forschungsf%C3%B6rderung/WKZ/Organic+far ming+in+Germany.pdf. Pal, L.A., 2011. Assessing incrementalism: formative assumptions, contemporary realities. Policy Soc. 30, 29–39. https://doi.org/10.1016/j.polsoc.2010.12.005. Piou, C., Berger, U., Grimm, V., 2009. Proposing an information criterion for individualbased models developed in a pattern-oriented modelling framework. Ecol. Modell. 220, 1957–1967. https://doi.org/10.1016/j.ecolmodel.2009.05.003. PLCopen, 2000. IEC 61131-7: Fuzzy Control Programming (edition 1.0 - 2000). Pollak, J., Riekmann, S.P., 2008. European administration: centralisation and fragmentation as means of polity-building? West. Eur. Polit. 31, 771–788. https:// doi.org/10.1080/01402380801906064. Princen, S., 2013. Punctuated equilibrium theory and the European Union. J. Eur. Public Policy 20, 854–870. https://doi.org/10.1080/13501763.2013.781822. Quinn, J.D., Reed, P.M., Keller, K., 2017. Direct policy search for robust multi-objective management of deeply uncertain socio-ecological tipping points. Environ. Modell. Softw. 92, 125–141. https://doi.org/10.1016/j.envsoft.2017.02.017. Ramchandani, P., Paich, M., Rao, A., 2017. Incorporating learning into decision making in agent based models. In: EPIA Conference on Artificial Intelligence. Cham. Springer International Publishing, pp. 789–800. Russell, S.J., Norvig, P., 2010. Artificial Intelligence a Modern Approach. London. Seidl, I., Tisdell, C.A., 1999. Carrying capacity reconsidered: from Malthus’ population theory to cultural carrying capacity. Ecolog. Econ. 31, 395–408. https://doi.org/ 10.1016/S0921-8009(99)00063-4. Sharma, R., Gaur, P., Mittal, A.P., 2016. Design of two-layered fractional order fuzzy logic controllers applied to robotic manipulator with variable payload. Appl. Soft. Comput. 47, 565–576. https://doi.org/10.1016/j.asoc.2016.05.043. Shrestha, S., Dhakal, S., 2019. An assessment of potential synergies and trade-offs between climate mitigation and adaptation policies of Nepal. J. Environ. Manage. 235, 535–545. https://doi.org/10.1016/j.jenvman.2019.01.035. Stokes, A., Bocqu´ eho, G., Carrere, P., Salazar, R.C., Deconchat, M., Garcia, L., Gardarin, A., Gary, C., Gaucherel, C., Gueye, M., Hedde, M., Lescourret, F., Mao, Z., Qu´ erou, N., Rudi, G., Salles, J.-M., Soubeyran, R., Subervie, J., Vialatte, A., Vinatier, F., Thomas, M., 2023. Services provided by multifunctional agroecosystems: questions, obstacles and solutions. Ecol. Eng. 191, 106949. https:// doi.org/10.1016/j.ecoleng.2023.106949. Sun, Z., Lorscheid, I., Millington, J.D., Lauf, S., Magliocca, N.R., Groeneveld, J., Balbi, S., Nolzen, H., Müller, B., Schulze, J., Buchmann, C.M., 2016. Simple or complicated agent-based models? A complicated issue. Environ. Modell. Softw. 86, 56–67. https://doi.org/10.1016/j.envsoft.2016.09.006. Torfing, J., 2009. Rethinking path dependence in public policy research. Criti. Policy Stud. 3, 70–83. https://doi.org/10.1080/19460170903158149. True, J.L., 2000. Avalanches and incrementalism: making policy and budgets in the United States. Am. Rev. Public Administr. 30, 3–18. https://doi.org/10.1177/ 02750740022064524. True, J.L., Jones, B.D., Baumgartner, F.R., 2019. Punctuated-equilibrium theory: Explaining Stability and Change in Public Policymaking. Theories of the Policy Process, Second Edition. Routledge, pp. 155–187. Watts, K., Whytock, R.C., Park, K.J., Fuentes-Montemayor, E., Macgregor, N.A., Duffield, S., McGowan, P.J.K., 2020. Ecological time lags and the journey towards conservation success. Nat. Ecol. Evolut. 4, 304–311. https://doi.org/10.1038/ s41559-019-1087-8. Wesselink, A., Colebatch, H., Pearce, W., 2014. Evidence and policy: discourses, meanings and practices. Policy Sci. 47, 339–344. https://doi.org/10.1007/s11077014-9209-2. Wlezien, C., 1995. The Public as Thermostat: dynamics of Preferences for Spending. Am. J. Pol. Sci. 39, 981–1000. https://doi.org/10.2307/2111666. Wolf, T., Nardelli, N., Shawe-Taylor, J., Perez-Ortiz, M., 2023. Can Reinforcement Learning Support Policy Makers? A Preliminary Study With Integrated Assessment Models. ArXiv preprint. https://doi.org/10.48550/arXiv.2312.06527. Ypma, T.J., 1995. Historical development of the Newton–Raphson method. SIAM Rev. 37, 531–551. https://doi.org/10.1137/1037125. Zahrnt, V., 2009. Public Money For Public goods: Winners and Losers from CAP Reform. JSTOR. https://www.jstor.org/stable/resrep00114. Zeng, Y., Brown, C., Raymond, J., Byari, M., Hotz, R., Rounsevell, M., 2024a. Exploring the Opportunities and Challenges of Using Large Language Models to Represent Institutional Agency in Land System Modelling. EGUsphere. https://doi.org/ 10.5194/egusphere-2024-449. Zeng, Y., Brown, C., Byari, M., Raymond, J., Schmitt, T., Rounsevell, M., 2024b. InsNetCRAFTY v1.0: Integrating institutional Network Dynamics Powered By Large Language Models With Land Use Change Simulation. EGUsphere. https://doi.org/ 10.5194/egusphere-2024-2661. Zeng, Y., 2024c. PID-fuzzy logic enodogenous institutional model. Zenodo. https://doi. org/10.5281/zenodo.13998435. Zeng, Y., Shi, Y., Shahbaz, M., Liu, Q., 2024d. Scenario-based policy representative exploration: a novel approach to analyzing policy portfolios and its application to low-carbon energy diffusion. Energy 296, 131202. https://doi.org/10.1016/j. energy.2024.131202. Zhao, A., McJeon, H., Cui, R., Cyrs, T., Feldmann, J., Iyer, G., 2022. An “All-In” Pathway To 2030: Transportation Sector Emissions Reduction Potential. Zheng, S., Trott, A., Srinivasa, S., Parkes, D.C., Socher, R., 2022. The AI Economist: taxation policy design via two-level deep multiagent reinforcement learning. Sci. Adv. 8, eabk2607. https://doi.org/10.1126/sciadv.abk2607. Y. Zeng et al. Ecological Modelling 501 (2025) 111032 22