Yield estimation of sugarcane based on agrometeorological-spectral models

Abstract

This work assesses the performance of a yield estimation model for sugarcane (Saccharum offcinarum) that uses spectral data in conjunction with yield estimated from an agrometeorological model. The test site was the sugarcane plantations of the Lençóis Paulista municipality in São Paulo State, Brazil. Production data for four crop years were analyzed. Yields (tons of stalks/ha) of the first crop year (1983/84) were regressed against spectral and agrometeorological data of that same year and used to predict the yield the following crop year, i.e., 1984/85. The models to predict the yield in subsequent years (up to 1987/88) were developed similarly, incorporating all previous years' data. The yield estimations obtained from these models explained 69%, 54%, and 50% of the yield variation in the 1984/85, 1985/86, and 1986/87 crop years, respectively. The accuracy of yield estimations based on spectral data only (vegetation index model) and agrometeorological data only (agrometeorological model) that were also investigated gave poorer results than the combined agrometeorological - spectral model.