Estimating Italian Ryegrass (Lolium multiflorum L.) Aboveground Biomass Using RGB Vegetation Index with UAV Remote Sensing and Its Accuracy Verification

Abstract

We estimated the aboveground biomass of Italian ryegrass (diploid) using the red green blue (RGB) vegetation index calculated from images captured by an unmanned aerial vehicle (UAV)-mounted RGB camera and verified its accuracy. The UAV acquired nadir images at 20 m above the ground, 80％ overlap, 70％ side lap, and white balance set to “AUTO” using an autopilot system. Orthomosaic images were used to calculate the vegetation index, and the average R, G, and B pixel values were calculated by specifying the sown survey area using the geographic information system software QGIS. From the obtained pixel values, the normalized red index (NRI), normalized green index, normalized blue index (NBI), red green ratio index (RGRI), green red vegetation index (GRVI), modified GRVI (MGRVI), visible atmospheric resistance index (VARI), green excess index, green leaf index (GLI), and RGB vegetation index. The measured aboveground biomass of Italian ryegrass was divided into calibration and verification sets, and a model was created using the calibration data, with aboveground biomass as the objective variable and each vegetation index as an explanatory variable. A significant correlation (P<0.01) was observed between the vegetation indices and aboveground biomass for all nine vegetation indices except for NBI. In particular, the vegetation indices that showed high correlations with aboveground biomass were MGRVI, VARI, RGRI, GLI, and GRVI. Of these, only the RGRI was negatively correlated with aboveground biomass. Using accuracy verification data for vegetation indices, excluding NBI, which had no significant correlation with aboveground biomass, we estimated aboveground biomass using a simple regression calibration model and compared the estimated aboveground biomass with the measurements. In particular, a high correlation was found between the aboveground biomass estimated from the NRI, RGRI, GRVI, MGRVI, VARI, and GLI and the measurements, and the estimated aboveground biomass was distributed near the line of the measured = estimated value. The RMSEE of the estimated aboveground biomass in the regression equation between the estimated and measured values was the smallest for VARI (0.878 kg/m²), followed by RGRI (0.885 kg/m²), MGRVI (0.886 kg/m²), and GRVI (0.895 kg/m²). These results suggest that VARI, RGRI, MGRVI, and GRVI, which can be estimated with RMSEE values of 0.878-0.895 kg/m², may be suitable for use in estimating aboveground biomass.

Journal of Warm Regional Society of Animal Science, Japan 68（2）: 97-103, 2025