In situ remote sensing as a strategy to predict cotton seed yield

Abstract

Crop harvest scheduling and profits and losses predications require strategies that estimate crop yield. This work aimed to investigate the contribution of phenological variables using path analysis and remote sensing techniques on cotton boll yield and to generate a model using decision trees that help predict cotton boll yield. The sampling field was installed in Chapadão do Céu, in an area of 90 ha. The following phenological variables were evaluated at 30 sample points: plant height at 26, 39, 51, 68, 82, 107, 128, and 185 days after emergence (DAE); number of floral buds at 68, 81, 107, 128, and 185 DAE; number of bolls at 185 DAE; Rededge vegetation index at 23, 35, 53, 91, and 168 DAE; and cotton boll yield. The main variables that can be used to predict cotton boll yield are the number of floral buds (at 107 days after emergence) and the Rededge vegetation index (at 53 and 91 days after emergence). To obtain higher cotton boll yields, the Rededge vegetation index must be greater than 39 at 53 days after emergence, and the plant must present at least 14 floral buds at 107 days after emergence.

Keywords

• precision agriculture
• path analysis
• decision trees
• gossypium hirsutum