Pattern-oriented modelling as a novel way to verify and validate functional–structural plant models: a demonstration with the annual growth module of avocado

Abstract

Background and AimsFunctional–structural plant (FSP) models have been widely used to understand the complex interactions between plant architecture and underlying developmental mechanisms. However, to obtain evidence that a model captures these mechanisms correctly, a clear distinction must be made between model outputs used for calibration and thus verification, and outputs used for validation. In pattern-oriented modelling (POM), multiple verification patterns are used as filters for rejecting unrealistic model structures and parameter combinations, while a second, independent set of patterns is used for validation.MethodsTo test the potential of POM for FSP modelling, a model of avocado (Persea americana ‘Hass’) was developed. The model of shoot growth is based on a conceptual model, the annual growth module (AGM), and simulates photosynthesis and adaptive carbon allocation at the organ level. The model was first calibrated using a set of observed patterns from a published article. Then, for validation, model predictions were compared with a different set of empirical patterns from various field studies that were not used for calibration.Key ResultsAfter calibration, our model simultaneously reproduced multiple observed architectural patterns. The model then successfully predicted, without further calibration, the validation patterns. The model supports the hypothesis that carbon allocation can be modelled as being dependent on current organ biomass and sink strength of each organ type, and also predicted the observed developmental timing of the leaf sink–source transition stage.ConclusionsThese findings suggest that POM can help to improve the ‘structural realism’ of FSP models, i.e. the likelihood that a model reproduces observed patterns for the right reasons. Structural realism increases predictive power so that the response of an AGM to changing environmental conditions can be predicted. Accordingly, our FSP model provides a better but still parsimonious understanding of the mechanisms underlying known patterns of AGM growth.