Agriculture's impact on the global land system is immense, driven by both land expansion (a fivefold increase since the industrial era) and intensification/simplification. While these changes have boosted food production, they've often compromised vital ecosystem services (NCP) such as water and climate regulation, pest control, pollination, and soil protection. Climate change mitigation and biodiversity conservation efforts, aligned with the Paris Agreement and Kunming-Montreal Global Biodiversity Framework, could curb agricultural land expansion. However, growing demand and land competition might intensify existing agricultural landscapes, exacerbating landscape simplification and trade-offs between material appropriation and key regulating NCP. These NCP are crucial for sustained agricultural productivity and resilience against climate extremes and pests. Landscape complexity, particularly semi-natural habitat within agricultural landscapes, enhances biodiversity and NCP supply. Declines in this habitat negatively impact pollination, pest control, and crop yields. Historically, integrated land-system models have often neglected finer-scale drivers like landscape simplification, leading to potentially biased assessments and imbalanced policies. This study addresses these limitations by using a coupled modelling framework to assess how land competition and global land-use dynamics drive changes in material and key regulating NCP across multiple scales.

The literature extensively documents the detrimental effects of agricultural intensification and landscape simplification on biodiversity and ecosystem services. Studies highlight the consistent positive relationship between landscape complexity (compositional and configurational heterogeneity) and biodiversity, as well as the supply of key NCP such as pollination and pest control. The amount of semi-natural habitat in agricultural landscapes is a strong predictor of crop pollination success and pest control efficacy, with declines leading to reduced yields. Global studies emphasize the economic consequences of wild pollinator decline on crop production, particularly affecting low-income countries. The importance of landscape approaches to mitigate soil degradation, especially when combined with on-site measures, is also well-established. However, limitations in existing integrated land-system models, primarily their coarse spatial resolution, have hindered a comprehensive understanding of landscape-scale impacts. This paper addresses this gap by employing a high-resolution modelling approach to analyze the interactions between global-scale land use changes and local ecosystem services.

This study uses a coupled modelling framework combining the open-source land system model MAgPIE (Model of Agricultural Production and its Impact on the Environment) with the Spatial Economic Allocation Landscape Simulator (SEALS). MAgPIE simulates global land-system dynamics using a cost-optimization approach, integrating socio-economic and spatially explicit biophysical information from the LPJmL (Lund-Potsdam-Jena managed Land) model at a 0.5-degree resolution. SEALS downscales MAgPIE's land-cover changes to a 10-arc-second resolution (300m x 300m) based on adjacency, suitability, and eligibility. Four land-use scenarios are modeled: a business-as-usual (BAU) scenario and three alternative scenarios incorporating area-based conservation (PROTECT), carbon-focused land restoration (COACTION), and a landscape policy targeting semi-natural habitat conservation in farmed landscapes (MULTI). The scenarios are evaluated using indicators for material production, climate regulation, pollination sufficiency (using semi-natural habitat within pollinator foraging distances), and soil loss by water erosion (using the GloSEM platform). Pollination sufficiency serves as a proxy for wild pollination supply and configurational landscape heterogeneity, while soil loss indicates soil degradation. The NCP are categorized based on the IPBES conceptual framework.

The study projects considerable increases in global demand for land-based commodities by 2050, driven by dietary shifts and population growth. Food demand expands by 35%, with disproportionate increases in Sub-Saharan Africa, the Middle East & North Africa, and India. Demand for pollinator-mediated crops also increases significantly. Across all scenarios, land-use transitions occur due to growing demands and cost-effective carbon uptake. Cropland expansion is substantial in the BAU scenario but significantly reduced in scenarios with climate action (COACTION and MULTI). Area-based conservation alone (PROTECT) has only a minor effect on cropland expansion. Yield increases across all scenarios remain below historical rates. Losses of primary and secondary forests are substantial in scenarios without climate action, but drastically curbed in COACTION and MULTI. The MULTI scenario, which incorporates landscape policy, does not lead to additional cropland expansion or carbon losses. Analysis of landscape-scale changes reveals that enlarging protected areas and carbon-focused restoration alone are insufficient to reverse negative trends in landscape heterogeneity and NCP supply, like wild pollination. The BAU, PROTECT, and COACTION scenarios show increased landscape homogenization, while the MULTI scenario shows a reduction in cropland areas with low pollination sufficiency scores. Soil loss is highest in the BAU scenario and significantly reduced in COACTION and MULTI, indicating strong synergies between climate action and soil protection. While landscape policy (MULTI) leads to higher overall soil loss than COACTION, it significantly reduces soil loss in historical cropland areas.

The study's findings highlight the importance of integrated, multi-scale approaches to land management. While increased appropriation of material NCP is projected to meet growing demands, this could negatively impact regulating NCP if not addressed by dedicated interventions. Unilateral focus on material NCP appropriation leads to landscape simplification and soil degradation, threatening long-term productivity and resilience. The results demonstrate that area-based conservation and carbon policies can be enhanced by interventions promoting landscape heterogeneity. Conserving 20% semi-natural habitat in farmed landscapes globally is feasible without significant trade-offs regarding forest conversion, carbon losses, or yield increases. However, this requires cropland relocation, necessitating policies to prevent expansion into sensitive areas and compensate for distributional consequences. The study also emphasizes the need for productivity increases in remaining cropland to meet future demands, through closing yield gaps, efficient land management, and ecological intensification, while carefully considering potential repercussions for people, biodiversity, and NCP. Promoting off-farm ecosystem management and landscape heterogeneity could counteract negative trends and reduce reliance on external inputs.

This study emphasizes the need for integrated, multi-scale approaches to global land management that consider both material and regulating NCP. While ambitious climate and biodiversity targets can be met without compromising agricultural productivity, it requires a policy landscape that promotes landscape heterogeneity and addresses the distributional consequences of cropland relocation. Future research should focus on more detailed modelling of feedback mechanisms between NCP supply and crop productivity, and on the specific local context-dependent quality and effectiveness of semi-natural habitat conservation.

The study uses simplified representations of semi-natural habitats and does not consider the full range of soil degradation processes or the impact of second-generation bioenergy crops on landscape structure and pollination. Feedbacks between NCP supply and crop productivity are not fully incorporated due to modelling limitations. Land-cover allocation at the field scale is primarily driven by existing land-use patterns. The study's global scale analysis inherently involves simplifications and might not capture local context-specific nuances.