Answers are grounded in this paper and cite their sources.
In Sanya, a tropical coastal rice-growing region of China, rice mapping and biomass estimation are limited by frequent cloud cover and heterogeneous planting schedules. To address these challenges, we developed a rice-recognition framework for cloudy areas, and combined remote sensing images synthesized from multi-index median pixels with deep learning models to construct 15 models for recognizing rice planting area. The MD_4s model exhibited the best performance (precision: 0.81, recall: 0.86, and F1 score: 0.84). According to this model, the rice plantation area in Sanya increased from 1,803.956 to 2,437.223 ha between 2019 and 2024. We also used the Dynamic Growth Period Unification algorithm to estimate rice biomass: radar data, optical data and texture information were extracted during the heading stage, and the estimation model was developed by combining random forest (RF), support vector machine, gradient enhanced decision tree, and backpropagation neural network algorithms. The RF algorithm exhibited the best performance (R2: 0.746, RMSE: 290.531 g/m2, bias: 6.766 g/m2, and RRMSE: 23.850%). According to the model, rice biomass density and total biomass in Sanya increased from 1142.218 to 1177.410 g/m2, and 21120.266 to 28696.107 Mg between 2019 and 2024, respectively. These results demonstrate the potential of the proposed workflow for rice mapping and heading-stage biomass estimation in the Sanya region. Moreover, the findings indicate that government assessments have overestimated both rice planting area and yield. Future studies will combine drone and multisource remote sensing data to achieve early yield estimates.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: