Sensitivity of CERES-Maize simulated yields to uncertainty in soil properties and daily solar radiation

The sensitivity of CERES-Maize yield predictions to uncertainty in a set of soil-related parameters and solar radiation was evaluated in Pergamino, in the Argentine Pampas. Maize yields were simulated for Pergamino using a 31 years climatic record for a range of values of a group of important model input variables. The input variables considered (and the range evaluated) were: soil nitrogen content at sowing (from 20 to 80 Kg ha(-1)), soil organic matter content (from 1.75% to 4%), soil water storage capacity (from 150 to 200 mm), soil water content at sowing (from 50% to 100%) of total available water), soil infiltration curve number (from 76 to 82) and daily solar radiation (from -20% to 12% of the historical values). Then, a sensitivity analysis using a combination of mathematical and graphical approaches was performed to evaluate the model response to changes in the values of the input variables considered. Moreover, a simplified method based on the evaluation of the model sensitivity at extreme values of the input variables is proposed to evaluate the model non-linear responses with a reduced number of runs. Under the scenario evaluated, representative of the typical maize productions systems of the Argentine Pampas, the model results showed higher sensitivity to changes in radiation (normalized sensitivity were -0.69 and 0.45 for rainfed and irrigated conditions, respectively) than for the soil variables (normalized sensitivity ranged from 0.20 to 0.28). The CERES-Maize model was found to have similar sensitivity for the different soil inputs. Furthermore, some of the variables evaluated (soil curve number, soil water content at sowing and radiation under rainfed conditions) showed an important lion-linear response. (C) 2006 Elsevier Ltd. All rights reserved.