On the number of experiments required to calibrate a cultivar in a crop model: The case of CROPGRO-soybean

The conventional approach for calibrating/validating a crop model considers few to many experiments. However, few experiments could lead to higher uncertainties and a large number of experiments is financial and time consuming. The objectives of this research were to study the calibration uncertainties and to find out the optimum (cost-benefit) number of experiment required for a reliable CROPGRO-Soybean model calibration/validation. This study used 21 field experiments (BMX Potencia RR variety) sown in eight different locations of Southern Brazil between 2010 and 2014. The experiments were grouped in 4 classes (Individual experiment, season/year per location, experimental sites and all data together). The developmental average Relative Root Mean Square Error (RRMSE) decreased from 22.2% to 7.8% in individual swings to all data together group, respectively. Use only one experiment (individual sowings) to calibrate a crop model, could lead to a RRMSE of 28.4, 48, and 36% for R1, LAI and yield, respectively. In general, as the number of experiment used during the calibration increases, smaller is the RRMSE's. The group that showed the best cost -benefit during the calibration/validation was the group 2 (season/year per location). The use of 3 experiments (early, optimum and late sowing dates), will ensure a reliable calibration/validation keeping the research resources use efficiency. (C) 2017 Elsevier B.V. All rights reserved.