The objective of this work was to obtain a forecast model of coffee crop productivity in seven municipal districts of Minas Gerais State, Brazil. The harmonic analysis in Fourier series was applied on productivity series to each location. The obtained coefficients until the harmonic seventh were submitted to multiple linear regression in the first three principal components of a group of 33 inherent variables of coffee production. These variables were averages of 15 years corresponding to the same years of productivity and subdivided in four quarterly periods during the agricultural year (July to June). The model did not prove adequate, since the results presented errors of estimatives with great discrepancies evidencing the complexity of yield forecast of coffee crop.
agrometeorological modeling; principal components; water balance
