Analysing the behaviour of a hazelnut simulation model across growing environments via sensitivity analysis and automatic calibration

Hazelnut is an emerging tree crop in the global market and current production levels are unable to sustain the demand for fruits. Research efforts are needed to support private and public stakeholders with knowledge that can help improving the management of the agricultural systems, by enhancing yield levels and safeguarding environmental sustainability. This paper extends the application of a process-based hazelnut simulation model from the single orchard where it was initially calibrated to four orchards located in Italy, Chile and Georgia, using three-year experimental datasets (2015–2017). The model sensitivity to parameter changes was assessed twice (screening and quantitative analysis) to identify the most relevant parameters affecting the prediction of leaf area index, fruit yield and soil moisture. These parameters were automatically calibrated to maximize the model accuracy in predicting field data, by comparing model performances with original parameterization. According to current model structure, canopy expansion and plant dimensions accounted for most of the variability in leaf area index simulation. Photosynthesis and root water uptake played a prominent role in affecting the model predictions of yield and soil moisture, respectively. The automatic calibration increased the accuracy in predicting leaf area index (average RRMSE = 19.3%), fruit yield (average RRMSE = 16.3%) and soil moisture (average RRMSE = 12.5%). This study provides quantitative information to support calibration activities and insights for collection of new experimental data in the short-term, while laying the groundwork for the use of the model for yield forecasting activities as the medium goal.