Calibration, Validation and Improvement of a Process-based Crop Simulation Model for Hardneck Garlic (Allium sativum L.)

Abstract

Crop models are powerful tools for assessing climate impacts on crops, assisting breeding and crop management decisions, predicting yield, and providing information for economic and policy decision making. Calibration and validation of these models are essential steps to evaluate the performance of a model in order to bridge the gap between simulation and reality, while identifying the areas that require improvements. In this study, a process-based crop model developed for hardneck garlic (Allium sativum L.) was tested for the accuracy of predicting the phenological, morphological and physiological processes to simulate growth and development of a whole plant. Planting date, cultivar, and nitrogen level differences were applied in order to test the model performance under a range of storage and environmental conditions, and also to explore the phenotypic variability within the species. Leaf development phenology, leaf area accumulation, carbon partitioning, biomass, and yield, as well as leaf photosynthesis were measured and compared with model outputs to determine model performance. Modifications were made in phenology, morphology and photosynthesis modules to incorporate the effects of planting date, cultivar and nutrient levels, in an attempt to improve model simulations while obtaining a reasonable level of parsimony. Field results showed that plant growth differed phenologically and morphologically amongst planting dates and cultivars. Differences in leaf development timing and final leaf number, along with changes in leaf length and width led to an overall inaccuracy in the simulation of leaf area accumulation, biomass gain, and final crop yield. Temperature-based phenological parameters were calibrated, and leaf area accumulation algorithms were restructured to incorporate the difference observed amongst planting date and cultivar groups. Model modifications resulted in improved model performance in predicting phenology, leaf area development and total biomass accumulation. Further iterations of algorithm development, calibration, and testing will be needed to improve and expand its capability to include additional environmental and agricultural management factors. The calibration and validation process in relation to planting dates, cultivar, and nitrogen levels in this study was the first attempt to evaluate and improve the model performance at the whole-plant level, which provided a better understanding of the model, pinpointing the strengths and weaknesses and showing areas for improvement.