Agricultural landscapes face a difficult balancing act. They must reliably produce food for growing populations, remain economically viable for farmers and rural communities, and sustain the ecological processes that underpin long-term productivity. Decades of agricultural intensification have simplified these landscapes, shrinking semi-natural habitats and fragmenting the ecological networks that deliver pollination, natural pest control, nutrient cycling, and hydrological regulation. When habitats become isolated or sparse, the ecological flows that support agricultural production diminish, threatening yield stability and resilience. A new study published in npj Sustainable Agriculture presents an integrated modeling framework that links farm-level ecological intensification, spatial habitat connectivity optimization, and a Bayesian search over realistic agri-environmental policies, applied to a real agricultural landscape in Manitoba, Canada.
The research team, led by Kevin Bradley Dsouza, Graham Alexander Watt, Yuri Leonenko, and Juan Moreno-Cruz, set out to solve a problem that has long frustrated both ecologists and economists. Ecological intensification, a strategy that seeks to enhance agricultural output by relying more heavily on ecological processes and less on synthetic inputs, is inherently a farm-level endeavor. Farmers adopt habitat restoration or biodiversity-friendly practices based on their private assessments of costs and expected gains. Even where such practices are profitable for many farms, the resulting habitat configuration often remains patchy and insufficient to sustain ecological connectivity. The cumulative landscape-level effects of these decentralized decisions therefore do not necessarily align with collective ecological goals.
The framework operates in three sequential layers. The first is a farm-scale ecological intensification model that determines economically optimal fractions for two spatially explicit interventions: enhancement of field margins, which allocates portions of field edges to semi-natural vegetation while maintaining crop production, and habitat conversion, which represents full restoration of selected cropland patches. These interventions differ in opportunity cost and ecological effect, and both have been shown empirically to strengthen pollination and pest control services. The model assumes farmers maximize net present value of profits over a planning horizon, incorporating revenues, input costs, ecological service gains, and opportunity costs associated with converting cropland to habitat. Yield per unit area is enhanced by pollination and pest suppression services that decay with distance, meaning the benefits of an intervention on one plot spill over to neighboring plots only within effective dispersal ranges.
When the researchers applied this farm-level model to the Manitoba study region, they found that optimal intervention strategies were crop-dependent. Canola and rapeseed plots distinctly favored margin interventions, showing the highest mean allocation by a large margin, driven by substantial ecosystem service benefits and high market value. Habitat conversion, while generally implemented at lower levels than margins, was most prominent in soybean plots at a mean of approximately 0.055 and corn plots at approximately 0.05. Other crops such as barley, oats, and spring wheat received minimal allocations of either intervention type. Economic factors, particularly maintenance costs, significantly influenced the overall strategy. Habitat conversion was associated with higher ratios of agricultural maintenance cost relative to habitat maintenance cost, and correlation analysis showed habitat conversion levels were positively driven by higher agricultural maintenance costs and negatively by higher habitat maintenance costs, though these correlations were weak.
Despite these local ecological and economic gains, the ecologically intensified landscape remained insufficient to generate meaningful connectivity at the regional scale. Habitat patches tended to emerge in isolated clusters or short linear segments, increasing ecological function around individual fields but leaving large gaps between major habitat nodes. This outcome underscores a core challenge of sustainable landscape design: privately optimal decisions often generate ecologically beneficial changes locally but are rarely sufficient to transform the structural connectivity of the broader landscape. The second layer of the framework addresses this gap directly. It uses a spatial optimization approach grounded in graph theory to identify where habitat restoration produces the greatest gains in the Integral Index of Connectivity, a metric widely used in landscape ecology that captures both habitat availability and the configuration of dispersal pathways.
The connectivity optimization produced a markedly different landscape configuration. When permitted to allocate additional habitat subject to the constraint that no farm’s profitability fell below an economically acceptable threshold relative to its ecological intensification baseline, the resulting pattern exhibited far greater ecological coherence. Compared to merely repositioning existing interventions, connectivity optimization yielded substantially higher connectivity scores, with the mean Integral Index of Connectivity increasing from 6.3 times ten to the third power after repositioning to 7.9 times ten to the third power after optimization. The optimization frequently identified narrow corridors and bridging patches in areas where large habitat clusters were separated by modest distances. These bridging patches, often less than a hectare in size, played an outsized role in shaping network structure. Restoring them lowered the effective distance between clusters to within the dispersal thresholds of key pollinator and natural enemy groups, thereby enabling multi-step ecological movement across previously fragmented regions.
The researchers identified what they call connectivity hubs, defined as agricultural plots ranking within the top ten percent by aggregated betweenness centrality, a metric that quantifies the extent to which a spatial unit lies along ecological pathways connecting habitat patches. Hub plots tended to have higher median yields, more complex shapes, and more neighbors. The authors suggest that plots with initially higher base yields are better able to economically sustain connectivity-enhancing interventions, and that the yield boost from ecosystem services on these high-yield plots helps ensure farm profitability remains within acceptable limits. Crucially, all connectivity improvements were achieved without violating the economic feasibility constraints. No farm in the optimized configuration lost more than the allowable margin of net present value relative to its baseline. This demonstrates that transformative ecological outcomes need not require large-scale cropland retirement or substantial economic sacrifice.
The third layer tackles the policy question. Because farmers respond to policy incentives rather than to global ecological objectives, the framework uses Bayesian optimization, a computationally efficient search strategy employing a Gaussian process surrogate model, to find the parameters of non-spatially explicit, farm-level policy instruments. These instruments include establishment subsidies, maintenance payments, eco-premiums, and habitat requirements. Each candidate policy is evaluated by re-solving the farm-level optimization under modified profit structures and measuring how closely the resulting landscape connectivity and farm profitability approach the theoretical benchmark. Significant improvements occurred within the first approximately fifty evaluations, after which the rate of improvement slowed, indicating the search likely converged toward a near-optimal solution within the defined parameter space.
The policy search revealed that there is no single universal solution. Many Bayesian-optimized policies achieved average connectivity scores comparable to or even exceeding the optimized baseline, and these ecological gains were often accompanied by higher average farm net present values, though the authors caution that some of these policies may break farm-level profitability constraints or impose unreasonable costs. Effective designs often involved moderate subsidy levels for intervention establishment and maintenance, with medians for various factors typically ranging from 0.2 to 0.3 within their 0.0 to 0.5 search spaces. However, the mandate parameter specifying the minimum fraction of margin adjacent to existing habitats consistently skewed toward lower values, with a median around 0.1 in successful policies, suggesting gentle nudges can be highly effective. Assigning a higher weight to connectivity shifted the optimization toward policies with stricter mandates or higher incentive payments, while prioritizing economic stability favored voluntary, lower-cost interventions that minimize disruption to farm operations, potentially at the expense of maximal landscape cohesion.
The authors acknowledge several limitations. The ecological intensification component captures only two major service pathways, pollination and natural pest control, and does not explicitly model other ecosystem functions such as soil fertility, hydrological regulation, or carbon storage. The connectivity metric is based on structural connectivity rather than species-specific functional connectivity, which may diverge for species with complex behavioral responses to landscape structure. The model also assumes farms respond deterministically to economic incentives, whereas behavioral studies show that adoption of conservation practices is often influenced by risk preferences, social norms, and perceptions of environmental stewardship. Furthermore, the framework captures static connectivity patterns, whereas real-world ecological networks are dynamic, with habitat maturation, crop rotation, and climate-driven range shifts altering connectivity priorities over time. Despite these caveats, the framework offers a flexible, adaptable tool for designing context-specific interventions, applicable to diverse governance structures ranging from farmer cooperatives to regional planning authorities, and demonstrates that coordinated restoration can transform fragmented agricultural mosaics into more resilient, multifunctional landscapes.
Subject of Research: An integrated multi-scale framework linking farm economics, landscape connectivity optimization, and agri-environmental policy search
Article Title: An integrated multi-scale framework for economically feasible landscape connectivity in working agricultural systems
Article References: Dsouza, K. B., Watt, G. A., Leonenko, Y., & Moreno-Cruz, J. (2026). An integrated multi-scale framework for economically feasible landscape connectivity in working agricultural systems. npj Sustainable Agriculture, 4(1), Article 89. https://doi.org/10.1038/s44264-026-00166-5
Image Credits: AI Generated
DOI: 10.1038/s44264-026-00166-5
Keywords: landscape connectivity, ecological intensification, agricultural landscapes, habitat restoration, Bayesian optimization, agri-environmental policy, pollination, pest control, ecosystem services, Manitoba, farm profitability, stepping stones
News Source: Alan Morgan. (November 5, 2026). Strategic habitat placement could connect fragmented farms without hurting profits. Scienmag.
