Prediction of the root anchorage of native young plants using Bayesian inference

The maximum vertical uproot resistance (Pmax) of a plant can be used to indicate its stability. Attempts were made to predict the Pmax of a plant empirically from some of its size parameters. In practice, an infinite number of empirical models with different complexities can be formulated. This study presents a Bayesian model class selection method to evaluate the plausibility of each empirical model among a list of model candidates. The models were ranked to identify the most plausible one. A database of vertical uproot resistance of four shrubs and trees native to Hong Kong was first compiled by performing field uproot tests. The plant size parameters including height, basal diameter, canopy size, and above-ground dry weight, were measured before the uprooting. Second, the mathematical formulation of the Bayesian model class selection method was presented. Using this method the most plausible model for each studied species or plant type was identified. The uncertainty of the model coefficients and therefore the credibility of the predictions were quantified. The selected models were used to predict the Pmax of the studied plants in some additional tests. It was found that the Pmax of the studied species could be reliably predicted from the established simple empirical formulas containing only the height and/or basal diameter of the plant as the dependent variables. The study paves the way for nondestructive and reliable prediction of plant anchorage.