Quantification of Approaching Wind Uncertainty in Flow over Realistic Plant Canopies

Abstract

Numerical simulations and in-situ measurements represent two important and synergistic pillars for the study of flow and transport in plant canopies. Due to model limitations and parameter uncertainty, the alignment of model predictions with actual observations is challenging in practice. The present work proposes a Bayesian uncertainty quantification (UQ) framework that estimates the approaching wind angle parameter for large-eddy simulation (LES) of flow in plant canopies by assimilating data from in-situ measurements. The framework is applied to LES of flow within and above realistic plant canopy, with plant area density derived from light detection and ranging measurements. Uncertainty on approaching wind direction is characterized via a Markov chain Monte Carlo procedure, and propagated through Monte Carlo sampling to wind speed and resolved Reynolds stresses. Given the substantial computational cost of LES, a surrogate model based on an exiguous number of LESs is used for flow simulations within the UQ framework. As a result of the analysis, the UQ solution is given by probability density functions of selected flow statistics at different heights. Profiles of mean ± standard deviation for the considered flow statistics exhibit excellent agreement with corresponding observations, proving that the proposed approach is able to calibrate the approaching wind angle parameter, and that the quantified uncertainty captures discrepancies between observations and model results. Overall, the present work highlights the potential of UQ to enhance predictions of exchange processes between vegetation canopy and atmosphere.