Evaluation of calibration strategies for rice modelling

In a previous study, the crop simulator CropSyst was evaluated against crop data collected on rice varieties grown in Northern Italy. The need of model re-parameterization became apparent after investigating new field data, as inconsistencies in the simulation of leaf area index emerged for Indica-type varieties. Key parameters (specific leaf area, stem-leaf partition, extinction coefficient, light-to-biomass conversion efficiency) derived from field measurements (respectively, 27 m2 kg-1, 3.6 m2 kg-1, 0.53, 3.2 g MJ-1) considerably differed from those previously obtained via calibration (39 m2 kg-1, 1.5 m2 kg-1, 0.50, 3.0 g MJ-1). Such new parameters are informative for suitable modelling of rice systems. The agreement between simulated and observed above ground biomasses was similar with both parameter sets: average general standard deviation = 25% (previous) and 26% (new); average modelling efficiency = 0.90 (previous) and 0.87 (new). Such comparisons demonstrate as the accumulation of aerial biomass in crop models can be depicted in different ways and reasonable estimations can be achieved by different pathways (not all acceptable). A check on parameters like the one performed here (field measurements versus calibrated parameters) is worth to give protection against spurious conclusions while indicating whether the parameterization is conceptually consistent and related to reality.