Coffee yield estimation by Landsat-8 imagery considering shading effects of planting row's orientation in center pivot

Abstract

Using spectral information, obtained from orbital sensor systems, combined with spectral indices, it's possible to characterize coffee trees conditions with their intrinsic characteristics, which can improve the imaging accuracy and optimize the crop monitoring tasks. However, within center pivot shape, a circular planting orientation is a feature that can difficult to determine coffee yield through remote sensing, resulting in inaccurate inferences about its spectral signature and its yield correlation. The objective of this study was to evaluate the correlation of different phenological stages spectral responses of coffee, under distinct apparent brightness conditions, with the yield sampled in field. The study was carried out on a center pivot irrigated area, located in Presidente Olegário municipality, Minas Gerais, Brazil. Yield data was obtained of 114 georeferenced sample points. Landsat-8 data, obtained between 05/25/2015 and 06/28/2016, were submitted to statistical analysis of Correlation and Multiple Linear Regression in conjunction with yield data. It was observed that the positional arrangement and the planting orientation significantly interfered in the yield estimation. A brightness difference was observed within the collateral area sectors, which raised the coffee plants spectral complexity. This demanded a conditional approach for data analyses and interpretation processes. Based on results, an evidence of good explanatory and predictive relationship with the coffee yield was found for Landsat-8 images when divergent brightness subsets were evaluated separately, raising the forecast accuracy from spectral data. It was concluded that the coffee phenological stages with the highest yield predictive potential are the “dormant bud” and “lead bead” stages.