Rice yield estimation using root length and biomass traits based on multivariate regression

Abstract

Understanding how root traits regulate rice yield is essential for improving productivity in environments where below-ground limitations restrict performance. This study estimated rice yield (RY) based on four structural root characteristics, root biomass (RB), root length (RL), root weight density (RWD), and root diameter (RD) measured across five sampling dates from 1 August to 11 September 2025, using power, logarithmic, linear, and exponential models. Strong positive relationships were detected for RB and RL, with the highest coefficients of determination reaching r² = 0.95 for RB and r² = 0.92 for RL. Conversely, RWD and RD showed consistent negative associations with RY across the evaluated sampling dates, with the lowest coefficients of determination reaching r² = 0.38 for RWD and r² = 0.51 for RD. Pearson correlations confirmed this pattern, revealing significant positive associations for RB (r = 0.92) and RL (r = 0.90), and strong negative correlations for RWD (r = −0.79) and RD (r = −0.78) at p < 0.01. PCA indicated that the first two components explained 90.40% of the total variance, with RB, RL, and RY clustering positively on PC1, while RWD and RD loaded negatively. A multiregression model using RB and RL achieved high predictive performance (R² = 0.91, RMSE = 0.43), and observed–predicted values aligned closely with the 1:1 line. These findings highlight the dominant contribution of fine-root development to yield formation and demonstrate the potential of root-based indicators for rice yield prediction and breeding applications.