A Boundary-Consistent and Topology-Sensitive Framework for Assessing Cropland Fragmentation Under Land-Cover Transitions: Evidence from Zhejiang, China

Monitoring cropland change requires attention to spatial configuration as well as total area. However, conventional fragmentation assessments can introduce artificial divisions at reporting boundaries and overlook how land-cover transitions reorganise existing patch networks. Using annual 30 m land-cover maps of Zhejiang, China, for 1990–2025, this paper developed a boundary-consistent cropland fragmentation index (BC-CLF) and related changes in the index to the topology of cropland gain and loss events. Conventional within-unit calculations overestimated BC-CLF by a median of 30.4% at the 1 km support. Although cropland area and fractional patch count declined, fragmentation increased because of greater spatial separation and edge exposure rather than patch proliferation. At the 8 km support, incorporating event topology improved prediction beyond transition quantity, direction, and source–destination type, increasing median R2 from 0.392 to 0.503 in spatially blocked cross-validation. The increment persisted after controlling for event abundance and size. Topology provided additional predictive information across all tested spatial supports, although the magnitude of improvement varied. By integrating boundary-consistent measurement with event-level structural information, the framework provides a basis for identifying changes in cropland configuration that area statistics alone cannot reveal.