Mapping drivers of grain and non-grain cropland transitions under rapid urbanization in the Pearl river delta using remote sensing and structural equation modelling

Clarifying spatiotemporal dynamics and driving mechanisms of paddy field conversion between grain and non-grain uses in the Pearl River Delta (PRD) is important for understanding agricultural sustainability in rapidly urbanizing regions. Frequent cloudy weather, fragmented farmland, and limited high-resolution data have constrained existing methods, and systematic analyses of cropland conversion pathways remain limited. We integrated multi-source remote sensing with a Random Forest model coupling optical and multi-temporal radar data to map paddy field conversions from 2016–2023 (OA: 93.92%; R2 > 0.88). Non-grain conversion areas for 2016–2018, 2018–2020, and 2020–2023 were 153,143 ha, 151,801 ha, and 135,219 ha, consistently exceeding grain re-conversion areas (134,716 ha, 117,762 ha, and 112,797 ha). Transitions were more active in peripheral than core areas. Non-grain conversion was statistically associated with agricultural labor (+1 ha/100 workers) and showed a negative association with fertilizer use (−0.06 ha/ton) and farmland aggregation (−431.55 ha/unit). Macro-level factors exerted indirect effects via labor allocation and land-use efficiency. Grain re-conversion showed strong associations with agricultural labor (+3 ha/100 workers) and farmland aggregation (+74.25 ha/unit), with weaker macro-level ties. These findings provide a technical pathway for cloudy, fragmented regions and reveals asymmetric transition patterns that may inform farmland regulation, non-grain conversion governance, and food security strategies.