Computational Models Growth & Development

Pattern-oriented modelling a useful tool for FSP models when experimental data is lacking

POM parameter estimation led to a model that converged on manual parameterization.

Functional-structural plant (FSP) models are used to aid in understanding the links between a plant’s physical architecture and the developmental mechanisms that create it. One of the key challenges of FSP modelling is parameter estimation, because in some cases, collecting data through field- or lab-based experiments is difficult or impossible, and parameter values must be established indirectly, through calibration. Pattern-oriented modelling (POM) is a means of calibration in which observed patterns are used to reject unrealistic parameter combinations. In this case, a pattern is any observation of non-random variation that can be assumed to contain information on the mechanism that produced it.

In a recent technical article published in Annals of Botany, lead author Ming Wang and colleagues investigated whether POM can make parameter estimation for FSP models more efficient and powerful than manual calibration, and whether it can reduce uncertainty when experimental data is unavailable. The authors used an FSP model of avocado branching architecture and tested POM parameterization against an existing manual parameterization to see if the two would converge.

Calibration with POM allowed the model to successfully reproduce the verification patterns, and even to “predict further independent validation patterns that were not used for model parameterization simultaneously,” write the authors. While POM calibration didn’t produce a single optimal parameter set, the pooled output, consisting of 22 parameter sets, was adequate and able to predict emergent system behaviours.

In fact, though one parameter set was similar to the manual calibration, the authors caution against using only the single best set. “Our study demonstrates that it would be risky to focus on just one single parameter set. The single set may make the model work perfectly under certain observed patterns, but when new patterns are observed, it may not result in the model producing the reliable outputs that match with the new observations.”

“The POM calibration approach allows FSP models to be developed in a timely manner without relying heavily on field or laboratory experiments, or on cumbersome manual calibration,” write the authors.