Abstract
In unmanned aircraft systems (UAS) remote sensing, flight altitude and pixel extraction size are important factors that determine the accuracy of soil physical and chemical property estimation. In this study, UAS multispectral imagery of bare and submerged soils were acquired at flight altitudes of 60 m and 140 m, the former with a ground sampling distance (GSD) of 5 cm/pixel and the latter with a GSD of 10 cm/pixel. Pixels were extracted from 4 sizes of buffer circles (radius 0.3 m, 0.5 m, 1.0 m and 1.5 m) to calculate the spectral index and the correlations of these indices with soil chemical properties were evaluated. Imagery taken at the 140 m altitude had higher correlations, suggesting that the lower GSD imagery could reduce the effect of interfering pixels such as crop residues and shadows. Although larger pixel extraction areas had higher correlations, there were some cases where the correlations were lower at the 1.5 m radius buffer circle. The multispectral band reflectance statistics showed that the data range increased from the 1.0 m radius buffer, suggesting that the correlation would be higher if the extraction areas included a variety of pixels. On the other hand, areas that were too large also included interfering pixels, resulting in lower correlations. It was considered that the optimal pixel extraction size was 1.0 m radius buffer for high accuracy of soil chemical estimation in practical use.
https://ror.org/01529vy56 
