UAV-multispectral and vegetation indices in soybean grain yield prediction based on in situ observation

Abstract

Remote sensing techniques using vegetation indices (VI's) have successfully enabled the accurate and timely monitoring of several crops. The purpose of this monitoring is to provide an advantage in terms of on-farm management decision making, crop marketing planning, and support for policy decision-making. We identified here which VI's can be used to predict soybean grain yield (GY) by using UAV (Unmanned Aerial Vehicle) and remote multispectral sensor. For this purpose, experiments were carried out in three sites during the 2017/2018 and 2018/2019 crop seasons in Brazil. The VI's were measured together with the plant stand assessment at 25 days after emergency (DAE) when the plants were at V4 phenological stage. The processing of the VI models was performed based on the imaging reflectance factor data performed in the field. For statistical data analyses, a correlation network was used to express the relationship between GY and VI's graphically. Path analysis was performed for identifying the cause-and-effect relationship between VI's and GY. Subsequently, a decision tree algorithm was generated considering GY as a dependent variable. At last, the relative deviation coefficient was used to illustrate the differences between the VI's in the construction of the decision tree and GY. Our results showed potential for predicting soybean yield based on UAV and the multispectral sensor coupled. SAVI and NDVI indices stood out for predicting yield, where the regions with the highest values of these indices can obtain the highest yield observed in the field, providing an advantage in management at the property level.