This deliverable provides a report on the agent-based models (ABMs) for each of the case studies, developed in the Work Package 4 (WP4) – Agent-Based Modelling and Analysis of BESTMAP. In particular, it includes a description of how the models can be accessed and which input data is needed. This document is accompanied by a description of each case study model in a structured form (see Appendix) which follows the ODD+D protocol (Müller et al., 2013). Deviations from the main processes as described in Milestone M6 (First versions of ABMs for CS) are discussed for each case study.

This Deliverable provides a roadmap to expansion of BESTMAP towards a operational pan-European modelling platform, as well as explore via pilot analyses several areas for improvement and future research. Considering new case studies, we analyse the locations where models parameterized in those regions can transfer to cover the most area. We conclude that in future case studies, they should be located in northern Spain, north-west Italy, central Italy, Montenegro/Albania, and Bulgaria. Testing if one can model water quality at the European scale, our modelling shows the NDR model (used in BESTMAP CS work) has generally good performance at EU scale, despite it being a rather simple process-based model. There is an overestimation of Nitrogen at low N, and underestimation of Phosphate at high P, which need to be considered in future work.

Model-based analyses can effectively contribute to investigating leverage points for sustainability transformations in agriculture. They allow for a systematic evaluation of policies under changing environmental, economic, or institutional conditions, and can be used to assess the effectiveness and efficiency of different policy designs. For analyzing agricultural systems, agent-based modeling is particularly useful because it can represent individual farmers—the crucial actors in land use systems—their interactions and emerging patterns at the landscape level. In order to provide policy-makers with relevant and accurate information, an adequate representation of farmers’ decision-making is essential. However, formalizing empirically observed farmers’ behavior into model rules is challenging, in particular when the observations are qualitative. With this article, we aim to guide modelers through the process of formalizing farmers’ decision-making based on empirical findings. First, we discuss which primary data collection designs are appropriate for inferring particular aspects of farmers’ behavior, focusing in particular on when a theory-driven design is helpful and when inductive approaches are needed. Second, we compile aspects that need to be covered in empirical data to best inform agent-based models. Finally, we present approaches for translating empirical findings into formalized decision rules. We underpin our discussion with model examples from the literature and our own model developed to represent farmers’ decision-making on the adoption of agri-environmental schemes in Europe. With this methodological contribution, we aim to help make agent-based models less stylized, thereby providing greater potential to support policy-makers in identifying leverage points for a sustainable transformation of agriculture.

This deliverable report provides an integration guide on how information gained in BESTMAP’s agent-based model can be used in the standard economic model to improve the assessment of agricultural policies in the European Union. First, the models used in the BESTMAP are explained. The integration guide discusses in detail the preconditions and challenges when linking agent-based models with standard economic models such as partial and general equilibrium models. As a result of an expert workshop, six challenges are identified. The report also presents suggestions on how to make use of the finding and presents a way forward to integrate the two types of models.

Agri-environment schemes (AES) are government-funded voluntary programs that incentivise farmers and land managers for environmental friendly farming practices. Understanding farmers’ decision-making process and its impact on AES adoption can aid policy makers in designing AES schemes that meet adoption goals and environmental targets. Farmers’ decision-making is complex and involves a range of social, behavioural, economic and ecological factors. In this paper, we present a spatially explicit agent-based model (ABM)—BESTMAP-ABM-UK that simulates farmers’ decision-making process, inclusive of farmers’ social, behavioural and economic factors, on adopting buffer strips, cover crops, grassland management and arable land conversion to grassland schemes in the UK. The model produces farmers’ AES adoption under varied AES scheme designs in term of the contract length, the offered payment level, the bureaucracy level and the required minimal area. We apply the Morris screening method to analyse the importance of the model parameters in a status quo scenario, in which current UK AES designs are used. The results show that the average accepted payments of farmers for buffer strips and grassland management and farmers’ intrinsic openness to buffer strips have the most significant impact on the farm adoption rate in the model.

Effective risk management is key to strengthening the resilience of the most vulnerable. Insurance products can help achieve this goal. Yet, in particular in low-income countries, not all can afford the regular premiums. Instead, informal risk-sharing within social networks plays a crucial role in protecting against unpredictable environmental shocks such as droughts and floods. However, this support may not reach households in need if income is heterogeneously distributed and poor households are not connected to those with sufficient resources to share. To study the determinants of vulnerability to extreme events when insurance is available but not affordable to all, we aggregate outcomes of an empirically informed agent-based model in a regression model. We show that not only a household's own financial situation is important for shock resilience, but also the household's position in the network and the financial situation of connected households. We demonstrate the transferability of our results by achieving high prediction accuracy for an empirical risk-sharing network of a village in the Philippines. The study demonstrates how model-based results can help detect vulnerable households. This can be used to develop a reliable identification tool to leverage external support and effectively target subsidies for different types of shocks.

This deliverable provides a General Framework for the BESTMAP Policy Impact Assessment Modelling (BESTMAP-PIAM) toolset. An update of the framework will be provided later in the project in Deliverable 2.4. The BESTMAP-PIAM is based on the notion of defining (a) a typology of agricultural systems, with one (or more) representative case study (CS) in each major system; (b) mapping all individual farms within the case study to a Farm System Archetype (FSA) typology; (c) model the adoption of agri-environmental schemes (AES) within the spatially-mapped FSA population using Agent Based Models (ABM), based on literature and a survey with sufficient representative sample in each FSA of each CS, to elucidate the non-monetary drivers underpinning AES adoption and the relative importance of financial and non-financial/social/identity drivers; (d) linking AES adoption to a set of biophysical, ecological and socio-economic impact models; (e) upscaling the CS level results to EU scale; (f) linking the outputs of these models to indicators developed for the post-2020 CAP output, result and impact reports; (g) visualizing outputs and providing a dashboard for policy makers to explore a range of policy scenarios, focusing on cost-effectiveness of different AES.

Sustainability challenges in socio-environmental systems (SES) are inherently multiscale, with global-level changes emerging from socio-environmental processes that operate across different spatial, temporal, and organisational scales. Models of SES therefore need to incorporate multiple scales, which requires sound methodologies for transferring information between scales. Due to the increasing global connectivity of SES, upscaling – increasing the extent or decreasing the resolution of a modelling study – is becoming progressively more important. However, upscaling in SES models has received less attention than in other fields (e.g., ecology or hydrology) and therefore remains a pressing challenge. To advance the understanding of upscaling in SES, we take three steps. First, we review existing upscaling approaches in SES as well as other disciplines. Second, we identify four main challenges that are particularly relevant to upscaling in SES: 1) heterogeneity, 2) interactions, 3) learning and adaptation, and 4) emergent phenomena. Third, we present an approach that facilitates the transfer of existing upscaling methods to SES, using two good practice examples from ecology. To describe and compare these methods, we propose a scheme of five general upscaling strategies. This scheme builds upon and unifies existing schemes and provides a standardised way to classify and represent existing as well as new upscaling methods. We demonstrate how the scheme can help to transparently present upscaling methods and uncover scaling assumptions, as well as to identify limits for the transfer of upscaling methods. We finish by pointing out research avenues on upscaling in SES to address the identified upscaling challenges.

Extreme weather conditions in the face of due to climate change often disproportionately affects the weakest members of society. Agricultural insurance programs that are specifically designed specifically for smallholders in developing countries are valuable tools that can help farmers to cope with the resulting risks. A broad range of methods including household surveys, experimental games, and agent-based models have been used to assess and improve the effectiveness of such climate insurance products. In addition Furthermore, process-based crop models have been used to derive suitable insurance indices. However, climate change raises specific socioeconomic andas well as environmental challenges that need to be considered when designing insurance schemes. We argue that, in light of these pressing challenges, some of the methodological approaches currently applied to study climate insurance reach their limits when applied independently. This has fundamental implications. On the one hand, not all undesired side effects of insurance can be detected and, on the other hand, insurance indices cannot be derived sufficiently well. We therefore advocate a sound combination of different methods, especially by linking empirical analyses and modelling, and underline the resulting potential with the help of stylized examples. Our study highlights how methodological synergies can make climate insurance products more effective in supporting the most vulnerable households, especially under changing climatic conditions.

Microinsurance is promoted as a valuable instrument for low-income households to buffer financial losses due to health or climate-related risks. However, apart from direct positive effects, such formal insurance schemes can have unintended side effects when insured households lower their contribution to traditional informal arrangements where risk is shared through private monetary support. Using a stylized agent-based model, we assess impacts of microinsurance on the resilience of those smallholders in a social network who cannot afford this financial instrument. We explicitly include the decision behavior regarding informal transfers. We find that the introduction of formal insurance can have negative side effects even if insured households are willing to contribute to informal risk arrangements. However, when many households are simultaneously affected by a shock, e.g. by droughts or floods, formal insurance is a valuable addition to informal risk-sharing. By explicitly taking into account long-term effects of short-term transfer decisions, our study allows to complement existing empirical research. The model results underline that new insurance programs have to be developed in close alignment with established risk-coping instruments. Only then can they be effective without weakening functioning aspects of informal risk management, which could lead to increased poverty.

Abstract Dynamic process‐based modelling is often proposed as a powerful tool to understand complex socio‐environmental problems and to provide sustainable solutions as it allows disentangling cause and effect of human behaviour and environmental dynamics. However, the impact of such models in decision‐making and to support policy‐making has so far been very limited. In this paper, we want to take a critical look at the reasons behind this situation and propose steps that need to be taken to change it. We investigate a number of good practice examples from fields where models have influenced policy‐making and management to identify the main aspects that promote or impede the application of these models. Specifically, we compare examples that differ in their extent to how explicitly they represent human behaviour as part of the model, ranging from purely environmental systems (including models for river management, honeybee colonies and animal diseases), where modelling techniques have long been established, to coupled socio‐environmental systems (including models for land use, fishery management and sustainable water use). We use these examples to synthesise four key factors for successful modelling for policy and management support in socio‐environmental systems. They cover (a) the specific requirements caused by modelling the human dimension, (b) the importance of data availability and accessibility, (c) essential elements of the partnership between modellers and decision‐makers and (d) insights related to characteristics of the decision process. For each of these aspects, we give recommendations specifically to modellers, decision‐makers or both to make the use of models for practice more effective. We argue that if all parties involved in the modelling and decision‐making process take into account these suggestions during their collaboration, the full potential that socio‐environmental modelling bears can increasingly unfold. A free Plain Language Summary can be found within the Supporting Information of this article.

Agricultural insurance is considered a promising instrument to manage climate risks and to enhance the food security of smallholder farmers. However, despite some positive evidence that insurance positively affects farmers' production strategies, consumption smoothing, asset protection, and asset recovery, the specific effect of insurance on farm households' dietary diversity is largely unexplored. Often, positive effects on dietary diversity are presumed through income gains that might arise from investment returns of profitable production activities and cash gains from payouts. We argue that there exist multiple other causal mechanisms through which insurance may even negatively influence farm households' dietary diversity. The current article elaborates these mechanisms and provides recommendations on ways to avoid unintended negative effects on dietary diversity which should be taken into account by governments and donors if they continue to further promote insurance.

Model-based analyses can effectively contribute to investigate leverage points for sustainability transformations in agriculture. They allow for a systematic assessment of policies under changing environmental, economic, or institutional conditions and can be used to evaluate the efficiency of different policy designs. This makes them an important tool for critically evaluating agricultural policies and shaping them appropriately to achieve the desired effect. For analyzing agricultural systems, agent-based modeling is particularly useful as it allows to represent individual farmers the crucial actors at the landscape level. This approach can explicitly incorporate farmer behavior to map, for example, the conditions for adopting sustainable practices that lead to more diverse agroecosystems.

Agent-based modelling (ABM) and social network analysis (SNA) are both valuable tools for exploring the impact of human interactions on a broad range of social and ecological patterns. Integrating these approaches offers unique opportunities to gain insights into human behaviour that neither the evaluation of social networks nor agent-based models alone can provide. There are many intriguing examples that demonstrate this potential, for instance in epidemiology, marketing or social dynamics. Based on an extensive literature review, we provide an overview on coupling ABM with SNA and evaluating the integrated approach. Building on this, we identify current shortcomings in the combination of the two methods. The greatest room for improvement is found with regard to (i) the consideration of the concept of social integration through networks, (ii) an increased use of the co-evolutionary character of social networks and embedded agents, and (iii) a systematic and quantitative model analysis focusing on the causal relationship between the agents and the network. Furthermore, we highlight the importance of a comprehensive and clearly structured model conceptualization and documentation. We synthesize our findings in guidelines that contain the main aspects to consider when integrating social networks into agent-based models.

Agent-based modelling (ABM) and social network analysis (SNA) are both valuable tools for exploring the impact of human interactions on a broad range of social and ecological patterns. Integrating these approaches offers unique opportunities to gain insights into human behaviour that neither the evaluation of social networks nor agent-based models alone can provide. There are many intriguing examples that demonstrate this potential, for instance in epidemiology, marketing or social dynamics. Based on an extensive literature review, we provide an overview on coupling ABM with SNA and evaluating the integrated approach. Building on this, we identify current shortcomings in the combination of the two methods. The greatest room for improvement is found with regard to (i) the consideration of the concept of social integration through networks, (ii) an increased use of the co-evolutionary character of social networks and embedded agents, and (iii) a systematic and quantitative model analysis focusing on the causal relationship between the agents and the network. Furthermore, we highlight the importance of a comprehensive and clearly structured model conceptualization and documentation. We synthesize our findings in guidelines that contain the main aspects to consider when integrating social networks into agent-based models.