Remote Sensing (Jan 2019)

Assessing Correlation of High-Resolution NDVI with Fertilizer Application Level and Yield of Rice and Wheat Crops using Small UAVs

  • Senlin Guan,
  • Koichiro Fukami,
  • Hitoshi Matsunaka,
  • Midori Okami,
  • Ryo Tanaka,
  • Hiroshi Nakano,
  • Tetsufumi Sakai,
  • Keiko Nakano,
  • Hideki Ohdan,
  • Kimiyasu Takahashi

DOI
https://doi.org/10.3390/rs11020112
Journal volume & issue
Vol. 11, no. 2
p. 112

Abstract

Read online

The aim of this study was to use small unmanned aerial vehicles (UAVs) for determining high-resolution normalized difference vegetation index (NDVI) values. Subsequently, these results were used to assess their correlations with fertilizer application levels and the yields of rice and wheat crops. For multispectral sensing, we flew two types of small UAVs (DJI Phantom 4 and DJI Phantom 4 Pro)—each equipped with a compact multispectral sensor (Parrot Sequoia). The information collected was composed of numerous RGB orthomosaic images as well as reflectance maps with spatial resolution greater than a ground sampling distance of 10.5 cm. From 223 UAV flight campaigns over 120 fields with a total area coverage of 77.48 ha, we determined that the highest efficiency for the UAV-based remote sensing measurement was approximately 19.8 ha per 10 min while flying 100 m above ground level. During image processing, we developed and used a batch image alignment algorithm—a program written in Python language–to calculate the NDVI values in experimental plots or fields in a batch of NDVI index maps. The color NDVI distribution maps of wide rice fields identified differences in stages of ripening and lodging-injury areas, which accorded with practical crop growth status from aboveground observation. For direct-seeded rice, variation in the grain yield was most closely related to that in the NDVI at the early reproductive and late ripening stages. For wheat, the NDVI values were highly correlated with the yield ( R 2 = 0.601–0.809) from the middle reproductive to the early ripening stages. Furthermore, using the NDVI values, it was possible to differentiate the levels of fertilizer application for both rice and wheat. These results indicate that the small UAV-derived NDVI values are effective for predicting yield and detecting fertilizer application levels during rice and wheat production.

Keywords