The objective of this study was to fit and validate a modified Jensen multiplicative model to estimate soybean grain yield in the State of Rio Grande do Sul, Brazil, under field conditions. The fitness was done using meteorological data from six weather stations located in the region of major production of this crop and data from averaged soybean grain yield over the whole state. The grain yield was obtained from official government statistics of IBGE (Instituto Brasileiro de Geografia e Estatística), from 1974/75 to 1994/95. The model showed a good fit, with determination coefficients varying from 0.86 for a complete model (November to April) to 0.75 for a reduced one (January to March). The model validation, done with independent data of 1995/96, 1996/97, 1997/98 e 1998/99, had a good performance, showing that water is the isolated factor that has the major influence on soybean grain yield definition in Rio Grande do Sul, and, therefore, could be incorporated into programs for predicting the crop harvest.
forecasting; models; water use; crop yield
