AGROMETEOROLOGICAL-SPECTRAL MODEL FOR ESTIMATING SUGARCANE PRODUCTIVITY IN BRAZILIAN SEMI-ARID

Abstract

This work aimed to develop an agrometeorological-spectral model, through a multiple linear regression, to estimate sugarcane productivity in the semi-arid region of Brazil. Annual agricultural yield data (2005/2006 to 2011/2012), monthly agrometeorological and spectral data (2005 to 2012) were used. In the calibration period of the model, the correlation between agrometeorological and spectral data in conformity with the real agricultural yield was the criterion chosen for the independent variables: irrigation plus rain precipitation, average air temperature, air vapor saturation deficit, and normalized difference vegetation index. In the calibration of the model, satisfactory results were observed with mean relative differences below 0.87% and an estimated standard error of 0.7806 tons of sugarcane in all crop years analyzed. In the model validation, the best performance was obtained for the crop year 2004/2005 compared to 2013/2014 and 2014/2015, what can be justified by the renewal of planting in this period. The model was adjusted through a correction factor and had its performance optimized in the 2013/2014 and 2014/2015 crop years. Multiple linear regression represents an excellent tool to be used in association with agrometeorological and spectral data for the estimation of agricultural productivity.