Canopy structure affects the estimation accuracy of the leaf area index when using vegetation index derived from spectral reflectance in rice

The leaf area index (LAI) is an important growth parameter for rice (Oryza sativa L.) crop. The vegetation index (VI) calculated from spectral reflectance can potentially be used for field-scale monitoring of LAI. Previous studies reported that VIs such as NDVI and EVI could be used to estimate LAI; however, their estimation accuracies often decreased during the later growth stages due to saturation of VIs. Furthermore, the differences in estimation accuracy among rice cultivars have not been well addressed. The objectives of this study were to identify suitable VIs for rice LAI estimation, and to address the differences in the estimation accuracies between rice cultivars using an optimal VI. Two rice cultivars with different canopy structures (LG5, an erect panicle [EP] cultivar and Koshihikari, a normal panicle [non-EP] cultivar) were grown under different fertility conditions. We continuously measured LAI and hyperspectral and multispectral reflectance using a plant canopy analyzer, handy spectroradiometer (HS), and unmanned aerial vehicle with a multispectral camera (UAV), respectively, from three weeks after transplanting to two weeks after heading. Seven VIs were calculated from hyperspectral and multispectral data to estimate LAI during the rice growth period. The results showed that Chlorophyll Vegetation Index yielded the most accurate LAI estimates among the calculated VIs, showing a linear relationship with LAI throughout the study period. The estimation accuracy did not decrease even when reproductive stage data were included. The results were consistent between spectral reflectance data obtained using the HS and UAV (R2 =0.78 for HS and R2 =0.85 for UAV). Moreover, the estimation accuracy of the EP cultivar was higher (R2 =0.93 for HS, R2 =0.95 for UAV) than that of the non-EP cultivar (R2 =0.71 for HS, R2 =0.82 for UAV). The EP cultivar showed uniform LAI vertical distribution across different nutrient environments, which might lead to better accuracy during the reproductive stage. The present study revealed that the estimation accuracy of rice LAI from VI was affected by differences in canopy structure. Further research is required to elucidate the relationship between canopy structure and estimation accuracy to establish a better method for LAI estimation.