The vulnerability of agricultural production systems to climate variation and extremes is increasingly being investigated with crop models, which poses the need of properly managing the (unavoidable) uncertainty associated with model estimates. Confalonieri et al. addressed the issue of assessing the uncertainty associated with model calibration by different users. Using five crop models of common use and different complexity (AquaCrop, DSSAT, CropSyst, STICS and WOFOST), four groups of five students were trained, arranged in a completely randomized block design with four replicates, and then mobilized in a trial-and-error adjustment of model parameters to match sets of observed biomass of maize and rapeseed crops. The results obtained in this study interestingly suggest that the uncertainty due to user’s subjectivity in model parameterization can be more relevant than the uncertainty inherent to model structure. This evidence is offered to the agricultural modelling community about the need of considering this type of uncertainty, especially in climate change impact studies.
Posted in Science briefs.