| Description: |
Accurate paddy rice mapping remains challenging in multi-season cropping systems, where overlapping phenological stages introduce classification uncertainties. While Synthetic Aperture Radar-based (SAR) methods provide cloud-independent monitoring, they often misinterpret phenological transitions, leading to confusion between cropping cycles and spectrally similar vegetation. In this study, we propose a Gaussian Time-weighted Growth Rate Index (GTGRI), a phenology-driven classification framework that integrates Growth Rate (GR) and Gaussian Time Decay (GTD) to enhance large-scale paddy rice mapping by using Sentinel-1 SAR time-series data. Compared to benchmark methods (Automated Rice Mapping using Synthetic Aperture Radar Flooding Signals (ARM-SARFS), Phenological-Knowledge-Independent (PKI), and SAR-based Paddy Rice Index (SPRI)) methods, the proposed GTGRI achieves high classification accuracy in single-season paddy rice regions like California Central Valley (CCV), U.S. (with an Overall Accuracy (OA) of 99.19% at 10 m and 97.03% at 30 m) and Goshogawara, Northern Japan (NJ) (with an OA of 94.66% at 10 m, 84.15% at 30 m). In double-season paddy rice regions, GTGRI effectively identifies dual cropping cycles, as demonstrated in Bago District, Southern Myanmar (SM) (with an OA of 76.51% and a F1 score of 0.83), highlighting its robustness in multi-cropping systems. By leveraging rice-specific phenological dynamics, GTGRI strengthens temporal separability between paddy rice and spectrally or structurally similar vegetation, such as lotus and reeds. This enhanced phenological discrimination also translates to improved performance in mangrove-dominated coastal zones, where GTGRI substantially mitigates misclassification induced by the overlapping seasonal backscatter coefficients of mangrove canopies. By dynamically adjusting to temporal variations, GTGRI overcomes the limitations of static classification models, enhancing consistency and transferability across diverse rice-growing landscapes. Its robust ... |