AimsPredicting the spatial distribution of root area remains challenging due to substantial intra- and inter-species variabilities. This study develops a new probabilistic framework for predicting cumulative root area profile of trees along depth by rigorously characterising model uncertainty.MethodsWe employed Bayesian methods to: (1) quantify the parameter uncertainty of the root distribution based on existing data; and (2) integrate new knowledge of the root distribution (i.e. rooting depth and maximum cumulative root area) to reduce the uncertainty in the prediction of root distribution. We developed single-level Bayesian models using exponential, Weibull, and power functions to characterise the lump-sum variability of fine root density and root diameter distribution. We also developed a hierarchical Bayesian model that characterised inter-trench and inter-species variabilities in cumulative root area profiles and enabled the incorporation of prior knowledge about maximum rooting depth and cumulative root area.ResultsThe power function-based model provided reasonable 95% confidence intervals and more conservative extrapolations than exponential or Weibull models. Inter-species variability significantly influenced variability in fine root density and root diameter distribution. Inter-trench variability of the cumulative root area profile exceeded inter-species variability.ConclusionsThe proposed Bayesian framework enables root distribution prediction for species lacking prior root distribution data, offering a robust tool for applying nature-based solution to mitigate shallow slope failures of forested terrains.