This study was carried out in the region of Palotina, state of Paraná (latitude 24º17'S, longitude 53º50'30", altitude 333 m), using the growth and development simulation model for soybean crop (Glycine Max (L.) Merrill), CROPGRO-Soybean, which was calibrated to the soybean cultivars CD 202, CD 204, CD 206 and CD 210, grown in the region of Palotina, state of Paraná. After calibration, the simulation of soybean yield was carried out for five different sowing dates and in 25 harvest periods, from 1974 to 1999. From the analysis carried out in the simulations, varying probability levels were determined for the actual occurrence of these productivities, for each of the simulated sowing times. From the five simulated sowing dates, October 1st presented the highest yields, but also the lowest. October 15th and November 1st were considered the best dates for soybean sowing, presenting an average for the four cultivars in the 25 simulations of 2,450 kg ha-1 and 2,303 kg ha-1, respectively. The yield probability analysis proved to be a great tool in decision-making as for the time of sowing of soybean. Well into the 50% level of yield probability is a good number for the farmer, in assisting in the decision of when to sow soybean in the region. Only the sowing time of December 1st presented productivity inferior to 2,000 kg ha-1 with probability levels of 50%.
simulation; CROPGRO-soybean; soybean