Navigating trade-offs in agricultural landscapes: Modelling transition pathways towards locally adapted livestock systems

Summary

Current agricultural systems face profound ecological imbalances, including soil degradation, biodiversity loss, water pollution, and disrupted nutrient cycles, all of which are intensified by climate change. Many of these imbalances are region-specific, shaped by local biophysical, socio-economic, and institutional contexts, such as watersheds, ecosystem-based units, or sub-national landscape units. As multifunctional and spatially heterogeneous systems, agricultural landscapes therefore represent critical leverage points for understanding these challenges and for developing targeted, context-specific solutions. Among disrupted nutrient cycles, phosphorus (P) exemplifies the challenge of managing a finite yet essential resource: excess P drives eutrophication and ecosystem degradation, while deficiencies constrain productivity in vulnerable regions. Beyond changes in consumption patterns, such as dietary shifts and reducing food waste, addressing these challenges calls for coordinated, landscape-level strategies, including water restoration, land use optimization, and improved soil, crop, and livestock management to balance P levels and enhance nutrient recycling. Livestock production is a major contributor to P surpluses in agricultural landscapes, especially in high-income economies such as Europe. While these surpluses pose risks to soil and water quality, livestock production also contributes to broader environmental pressures such as greenhouse gas emissions, water use, and biodiversity loss. At the same time, livestock provides essential functions in food production, rural livelihoods, and nutrient cycling. Ensuring the sustainability of livestock systems therefore requires balancing environmental, nutritional, and socio-economic objectives.
In this context, models play a vital role in guiding these sustainable transitions in agriculture by supporting the generation of systems, target, and transformation knowledge. They help to assess current landscape functions and trade-offs (systems knowledge), explore desirable futures through scenario analysis (target knowledge), and inform actionable pathways to achieve these futures (transformation knowledge). From field-level biophysical simulations to global food system models, a wide range of modelling approaches assist to generate this knowledge. Their integrative capacity makes them indispensable tools for assessing trade-offs, envisioning alternatives, and shaping transitions in agricultural landscapes.
Yet current modelling approaches for assessing agricultural landscapes face limitations in capturing their multifunctionality, integrating agronomic with other landscape functions, and representing linkages both within the landscape and to the broader food system. However, a comprehensive synthesis of these approaches is still lacking, leaving open which functions are typically assessed and how food system connections are incorporated. Furthermore, although reducing P surpluses in intensive livestock systems is recognized as a priority, the full range of viable reduction strategies, that balance environmental goals with production needs and farmer acceptance, remains poorly understood. Finally, practical, locally adapted pathways for transforming livestock systems over time to achieve significant P reductions in intensive livestock production are largely unexplored. This dissertation addresses these gaps by advancing model-based approaches for designing locally adapted agricultural systems and by illustrating pathways towards these desirable future states. In doing so, it contributes essential systems, target, and transformation knowledge to guide the future development of agricultural systems. The dissertation is structured as a cumulative thesis comprising three research articles.
The first article presents a systematic literature review of spatially explicit models used to assess agricultural landscapes. To structure the analysis, a theoretical framework defining key elements of agricultural landscape assessments was developed, enabling a systematic comparison of modelling approaches across studies. The analysis reveals that while biodiversity and ecosystem services such as Water conditions and Atmospheric composition/conditions are widely covered, they are rarely used in combination. Other services, like Pest and disease control, are largely absent. Moreover, most models lack a representation of return flows from the food system and rarely combine crop–livestock dynamics, limiting their ability to capture synergies or trade-offs comprehensively. This review thus identifies critical blind spots and highlights the need for more integrative modelling approaches to adequately support decision-making in multifunctional agricultural landscapes. Additionally, the proposed framework of key elements for agricultural landscape assessments offers a structured starting point for achieving a more comprehensive system understanding in future assessments.
The second article introduces LEAF.livestock, a novel dynamic livestock model developed to overcome some of these limitations. The model is applied to the livestock-intense catchment of Lake Sempach in Switzerland, where nutrient imbalances driven by high livestock densities and feed imports have led to excessive P accumulation. Nearly five million combinations of herd structures and management practices are simulated, covering changes in livestock density and composition, feed sourcing, and rearing pathways. Thereby, the analysis focuses on the consequences of these changes for P excretion, animal-sourced protein output, land use, and feed–food competition. No single scenario meets all environmental and production constraints, but when land-use restrictions are relaxed, over 514’000 viable and 7’174 Pareto-optimal scenarios emerge, each with distinct trade-offs: Pig-dominated scenarios achieve stronger P reduction and protein output but rely on more arable land and imports, while dairy-oriented scenarios are less productive but better align with regional land availability. These findings highlight the potential of multiple, locally adapted solutions rather than one global optimum.
Building on these results, the third article simulates transition pathways from the current livestock system to two contrasting target scenarios. Using a resistance-weighted optimization approach combined with the dynamic LEAF.livestock model, farm type and management changes were simulated from 2020 to 2060. Both transition pathways achieve the P reduction targets, but through different restructuring processes: the pathway towards the pig-dominated target scenario relies on farm exits and specialization, whereas the pathway towards the dairy-dominated target scenario emphasizes feeding plan changes with fewer exits. These findings highlight that while multiple routes can meet the same environmental and productivity objectives, each carries distinct social, structural, and policy implications, underlining the need for targeted support to ensure feasible and acceptable transitions.
Together, the three articles demonstrate how locally adapted modelling can inform livestock management and policy design by making trade-offs visible, identifying synergies, and supporting evidence-based planning. Focusing on the catchment of Lake Sempach, this dissertation contributes to the design of livestock transition pathways that reduce P excretion while maintaining food production. While the catchment serves as a detailed case study, the modelling approach developed in this dissertation offers broader relevance for agricultural landscapes facing similar challenges, providing a foundation for locally adapted decision-making and policy design.