Forecasting of the annual yield of Arabic coffee using water deficiency

Abstract

Abstract: The objective of this work was to develop agrometeorological models for the forecasting of the annual yields of Arabic coffee (Coffea arabica), using monthly water deficits (DEFs) during the coffee cycle, in important locations in the state of Minas Gerais, Brazil. For the construction of the models, a meteorological data set spanning of 18 years and multiple linear regressions were used. The models were calibrated in high- and low-yield seasons due to the high-biennial yields in Brazil. All calibrated models for high- and low-yield seasons were accurate and significant at 5% probability, with mean absolute percentage errors ≤2.9%. The minimum forecasting period for yield is six months for southern Minas Gerais and Cerrado Mineiro. In high-yield seasons, water deficits affect more the reproductive stage of coffee and, in low-yield seasons, they affect more the vegetative stage of the crop.

Keywords

• Coffea arabica
• agrometeorology
• climate
• forecasting
• model
• water balance