Examining the Applicability of UAV and Satellite Remote Sensing Data for Soybean Cultivation

概要

In this study, the suitability of remote sensing data from satellites and drones for soybean yield estimation was examined. The possibility of evaluation of field drainage using satellite remote sensing data was also examined. Test sites were on Shinchi Town in 2019 and 2020, and on Minamisoma City, Fukushima prefecture in 2021. The near-infrared band was extracted from Sentinel-2 satellite data obtained between March and June of 2019-2021 to evaluate the field drainage. Images observed by cameras mounted on a drone in 2019 and 2020, and by WorldView-2 satellite in 2021 were analyzed.

　Vegetation indices were calculated with acquired images. The vegetation indices used in this study were NDVI, GNDVI, VARI, and GCC. NDVI and GNDVI use near-infrared wavelengths in their calculations, while VARI and GCC were indices using only visible bands. Regression analysis was conducted with each vegetation indices and the soybean yields. Sentinel-2 data analysis showed that near-infrared reflectance from the bare soil on the target field decreased after obvious rainfalls. In addition, spatial variability of near-infrared reflectance in the field was observed. Sentinel-2 near-infrared reflectance images can be used as an indicator of the drainage of the field.

　Results of regression analysis between vegetation indices and the soybean yields showed that the correlations of VARI and GCC with soybean yield were comparable to those of NDVI and GNDVI. This result suggests the possibility to predict soybean yield using cameras that can observe only visible spectrum. Application of inexpensive digital cameras is expected to reduce the introduction cost for drone-based agricultural remote sensing.