• Japan's tea industry is shifting toward large-scale operations, necessitating more efficient harvest management. • Spectral reflectance and vegetation indices were derived from unmanned aerial vehicle-mounted multispectral sensing (UAV-MS) data. • Models were developed to estimate fresh leaf yield, total nitrogen (TN), and fiber contents. • The best yield models used RDVI and MTVI2 in the first season and DVI in the second. • MTVI2 was best for TN and fiber contents in the first season, while Green (G) was best in the second. • UAV-MS remote sensing achieved accuracy comparable to visual observation methods. Tea is a major economic crop in East Asia. In Japan, labor shortages in recent years have increased the demand for labor-saving harvest management. To determine the optimal harvest timing, it is necessary to accurately evaluate the fresh leaf yield and component concentrations that are directly related to the market price to maximize income. This study evaluated the utility of unmanned aerial vehicle-mounted multispectral sensors (UAV-MS) to estimate fresh leaf yield and quality-related components (total nitrogen TN and fiber content). The Field experiments were conducted at the Shizuoka Prefectural Tea Research Center during the first and second harvest seasons over three years (2021–2023). The UAV-MS collected spectral reflectance data from which the vegetation indices were derived. These were used to estimate the yield per 10 are and the chemical composition. The root mean square error (RMSE) values showed that the UAV-based estimates achieved an accuracy comparable to or exceeding that of traditional visual observation methods. In addition, key spectral bands important for prediction were identified. In both harvest seasons, eight vegetation indices involving near-infrared (NIR) and red (R) wavelengths were significant for yield estimation. For TN and fiber content, the same indices were useful in the first harvest season, whereas in the second harvest season, single-band data from the green (G), red edge (ED), and NIR wavelengths were more informative. These findings provide fundamental insights into the practical application of UAV-MS-based harvest management, which will improve estimation accuracy and promote upscaling to satellite-based observations.
Furuya et al. (Fri,) studied this question.