Estimation Accuracy of Growth Traits in Rice using Nondestructive Sensing

Abstract

Twenty-four treatments consisting of combinations of four nitrogen application levels and six rice (Oryza sativa L.) cultivars with high yield and/or high eating quality were examined. The accuracy was evaluated for the estimation of growth traits (leaf area index (LAI), number of tillers, dry weight, nitrogen content of leaf and nitrogen content of plant) by nondestructive sensing data (vegetation cover rate, normalized difference vegetation index (NDVI) and light interception rate). The relationships and estimation accuracy based on data set obtained at 24, 38, 54, 66 days after transplanting (DAT) were evaluated. The coefficient of determination (R²) of regression equation between vegetation cover rate and LAI, dry weight and nitrogen content of plant, NDVI and LAI, nitrogen content of leaf and nitrogen content of plant, light interception rate and dry weight were more than 0.8 from 24 DAT to 54 and/or 66 DAT (panicle initiation stage (PI)). In addition, the R² of regression equation between vegetation cover rate of handheld camera and dry weight and nitrogen content of plant at 24 DAT, and all nondestructive sensing data and growth traits except number of tillers at 24 and 38 DAT were more than 0.7. At PI, the R² of regression equation between NDVI of handheld sensor and growth traits except dry weight, and light interception rate and LAI were also more than 0.7. It was suggested that root mean square error (RMSE) tended to be large, while there were relationships between these nondestructive sensing data and growth traits.