Rapeseed Yield Estimation by Assimilating a Novel SAR-Retrieved Canopy Parameter Considering Plant Structure and Regional Characteristics

Real-time regional rapeseed yield estimation is essential for agricultural management, market stability, and public safety. However, traditional assimilation systems that use the leaf area index (LAI) as a state variable underestimate rapeseed yield because they ignore the photosynthetic contribution of siliques. To address this limitation, this article introduces the total photosynthetic area index (TPAI), which integrates both leaf and silique contributions, and incorporates it into an improved assimilation framework. A phenology-driven, layered microwave scattering model was developed by combining rapeseed canopy structure, regional characteristics, and microwave scattering theory. This model accurately captures the temporal dynamics of the TPAI and achieves strong inversion performance, with <inline-formula><tex-math notation="LaTeX">${{R}^{2}}$</tex-math></inline-formula> values of 0.743 (training) and 0.819 (validation). In the crop growth model, a localization approach that relies on the TPAI for parameter adjustment was introduced, improving the agreement between the observed and simulated total weight of storage organs (TWSO) from <inline-formula><tex-math notation="LaTeX">${{R}^{2}}$</tex-math></inline-formula> &#x003D; 0.500 (LAI-based) to 0.982. In the data assimilation component, key parameters were selected using the recursive feature elimination algorithm, and the yield was estimated using the ensemble Kalman filter with the TPAI as the state variable. Compared with LAI-based assimilation, the TPAI-based approach substantially improved TWSO estimation accuracy at the reginal scale, with <inline-formula><tex-math notation="LaTeX">${{R}^{2}}$</tex-math></inline-formula> &#x003D; 0.791, RMSE &#x003D; 1050.835 kg/ha, outperforming the LAI-based results ( <inline-formula><tex-math notation="LaTeX">${{R}^{2}}$</tex-math></inline-formula> &#x003D; 0.052, RMSE &#x003D; 2330.970 kg/ha). Overall, by accounting for plant structure and regional characteristics and by assimilating a novel SAR-derived canopy parameter, this TPAI-based framework effectively mitigates the yield underestimation inherent in LAI-driven methods.