Impacts of farmland trees on ecosystem services in intensive agricultural systems

This is the authors' abstract. We don't add key points for this paper.

Farmland trees, defined here as trees outside forests occurring within agricultural landscapes, provide critical ecosystem services in intensive agricultural landscapes, yet their regional distribution and ecological contributions remain poorly quantified because small, linear, and fragmented tree elements are difficult to map with conventional remote sensing. This study focuses on the intensively cultivated plains of Northeast China (∼469,000 km 2 ) and proposes a hierarchical classification framework integrating object-based segmentation and spatial semantics to produce a fine-scale map of farmland trees using multi-temporal Sentinel-1 and Sentinel-2 data, and to assess their associations with ecosystem services. The framework first generates a vegetation mask based on time-series vegetation-frequency analysis, then performs vegetation-type classification at the object level, and finally refines the identification and separation of farmland trees embedded within forested areas by incorporating morphological features and their adjacency relationships with farmland. The method achieves an overall accuracy of 81% (Kappa = 0.62), estimating total farmland tree area at 2033.62 km 2 . Furthermore, segmented linear regression was used to quantify the nonlinear relationships between farmland-tree density and key ecosystem services. Water yield was positively associated with density up to about 30%, after which the increase weakened. Sandstorm prevention was positively associated with density up to approximately 40%, after which the increase accelerated. Soil retention showed no significant association below approximately 15% density and a declining tendency at higher density levels. Overall, this framework provides methodological support for scalable and interpretable mapping of farmland trees in large-scale, heterogeneous farmland landscapes, and by identifying density thresholds for service responses, it offers a scientific basis for sustainable land management balancing agricultural productivity and ecological benefit.