UAV-Enabled Neural Network Classification of Convective Flow Structures for Energy Flux Analysis in the Atmospheric Boundary Layer

The identification and tracking of atmospheric flow structures could improve our understanding of flux balances in the atmosphere; this may help address the long-standing energy balance closure problem. The Eddy Covariance method, commonly used to measure surface fluxes, often lacks accuracy under varying conditions, particularly during daytime convection, which is the focus of this study. The present study has been conceived with the objective of developing a real- time algorithm capable of identifying and tracking flow structures, such as updrafts, based on UAV measurements. This algorithm would flag relevant structures in the vicinity of the measurement tower and assess their impact on the energy balance closure. Figure 1 shows a structure classification over an area of interest; updraft regions, in red, are clearly identified. Firstly, a classification neural network (NN) was trained using an extensive dataset from Large Eddy Simulations (LES) of convective boundary layers with various atmospheric conditions derived from real data from the ICOS tower in Lonzée, Belgium. This NN has exhibited strong performance, reaching 82% accuracy in identifying structures according to the Park et al. 1 classification, even under unseen atmospheric conditions during the training process. Initially, the NN functions exclusively with point-wise data. However, its accuracy drops near the ground, where significant heterogeneity occurs,and also at the border between two classes. To address this, the NN is retrained, incorporating spatial information along UAV trajectories, capturing gradients in the measurements. Several flight paths, including scanning and orbit-like trajectories, are analyzed to improve accuracy. Finally, the NN will be applied to real UAV data over the Lonzée site and compared with LiDAR measurements to assess accuracy. Results from both the numerical study and field data will be presented. 1 Park, S., P . Gentine, K. Schneider, and M. Farge, 2016: Coherent Structures in the Boundary and Cloud Layers: Role of Updrafts, Subsiding Shells, and Environmental Subsidence. J. Atmos. Sci., 73, 1789–1814,https://doi.org/10.1175/JAS-D-15-0240.1.