Early forecasting and spatial typologies indicate rice yield growth potential in Yangtze River Delta

With global population growth and limited arable land, improving crop yields has become a critical pathway to ensuring agricultural sustainability. Enhancing agricultural productivity fundamentally depends on accurate and timely estimation of crop yields and yield gaps (YGs). Although short-term yield forecasting has advanced, long-term early warning frameworks remain scarce. Moreover, reliance on post-harvest yield statistics and static zonation in YG analysis limits the spatiotemporal sensitivity required for forward-looking agricultural management and for alleviating food supply pressures. To address these challenges, we propose the YS-RYG-IMF framework, which integrates six-month-ahead yield prediction with dynamic YG identification to establish a closed-loop pathway from early evaluation to differentiated intervention. The framework comprises three modules: (1) rice yield prediction six months before maturity using the BC-HRF model, which incorporates climate, soil, vegetation, and topographic variables and mitigates regression-to-the-mean effects; (2) dynamic identification of the annual attainable yield (Ya, 95th percentile) through ISODATA clustering, while YG dynamics are quantified through a moving-window linear regression (MWLR) approach; and (3) classification of YG trend into “steady growth”, “ceiling pressure”, and “stalled floor” types via a three-element typology, followed by construction of a regional priority map for yield improvement. Applied to the Yangtze River Delta (2006–2021), the results showed that the BC-HRF model achieved effective early yield forecasting (R 2 =0.65, RMSE=700 kg ha −1 ), although uncertainties remained under extreme climatic conditions. Overall, regional rice yields increased steadily, while northern Jiangsu and northern Anhui exhibited slowing or declining trends. The “ceiling pressure” and “stalled floor” types showed increasing north–south spatial heterogeneity, suggesting that yield improvement potential is more constrained in southern regions. Priority zoning analysis indicated that 71.4% of Jiangsu was classified as first-priority areas, compared with 17.9% in Zhejiang and 10.9% in Shanghai, revealing pronounced interregional disparities in yield potential and management requirements. Further OLS regression results demonstrated that climate conditions, cultivation systems, and socioeconomic factors exerted differentiated driving effects across provinces. In summary, the YS-RYG-IMF framework links prediction, classification, and zoning to provide both a theoretical foundation and methodological support for early warning and differentiated management of agricultural systems under climate change, with transferability and policy relevance. • YS-RYG-IMF provides a spatiotemporal indicator system for early yield risk detection. • Dynamic yield gap typologies reflect long-term yield sustainability potential. • Yield gaps are classified into steady growth, ceiling pressure, and stalled floor types. • Priority regions for sustainable yield improvement are mapped for targeted action. • Climate, management, and socioeconomic factors drive province-specific YGs.