Mapping diverse rice cropping patterns with limited samples by learning phenology-aligned temporal similarity from multi-source remote sensing data

Accurate mapping spatial distribution of diverse rice cropping patterns (RCPs) is essential for agricultural monitoring and food security, yet remains difficult in fragmented, cloud-prone regions due to the limited availability of temporally transferable, spatially representative, and category-balanced training data, especially for historical mapping. We propose AutoRCP, an automated framework that generates representative and transferable training samples and enables long-term RCPs classification. Thirty-meter NDVI and LSWI time series were reconstructed via the Gap-Filling Savitzky–Golay (GF-SG) algorithm, from which cropping intensity (CI) was derived with an overall accuracy of 87.87%. The derived CI was mainly used as a prior constraint for constructing high-confidence single- and double-cropping sample pools by integrating it with APRA500 and ChinaCropPhen1km. Based on these sample pools, representative samples for rice monoculture (RM), double rice (DR), rice-upland rotations (RUR), ratoon rice (RR), and rice-crayfish (RCF) were generated using K-means clustering and phenology-window DTW (PWDTW), with thresholding ensuring intra-class consistency and inter-class separability. Quantitative evaluation showed close agreement between AutoRCP samples and field observations (high SCS, low SBD), confirming their suitability for classification. Using these samples, the 2021 map achieved overall accuracy (OA)=84.98%, with producers’ accuracy (PA) and users’ accuracy (UA) above 80% for most RCPs. County-level comparisons against census data indicated strong concordance for single-season rice ( R 2 =0.82-0.85, RMSE=13.24-16.49 kha) and significant correlations for double rice ( R 2 =0.68-0.77). From 2000-2021, RM remained dominant, RUR expanded, DR declined, and RR/RCF grew rapidly after 2015. Overall, AutoRCP offers a scalable path to reliable RCPs mapping, supporting sustainable agricultural management and food-security policy.