The identification and tracking of atmospheric flow structures can improve our understanding of complex atmospheric phenomena, including the long-standing energy balance closure problem. Traditionally, the Eddy Covariance method is used to measure surface fluxes, but its accuracy is limited under certain conditions, particularly daytime convection. This study develops a real-time algorithm to identify and track flow structures, such as updrafts, using UAV-based measurements. By flagging structures near flux towers, the algorithm can evaluate their impact on energy balance closure. Figure 1 illustrates the classification of flow structures over the area of interest, with updrafts highlighted in red. A classification neural network was trained on over 100 million samples from Large Eddy Simulation (LES) of convective boundary layers, spanning a range of atmospheric conditions (geostrophic wind: 1–4.5 m/s, mean temperature: 286.15–299.15 K, surface heat flux: 20–110 W/m²). These conditions were derived from ICOS tower data in Lonzée, Belgium. The network achieved 82% accuracy in identifying structures based on the Park et al. 1 classification, even under unseen conditions. Initially, the network operated on point-wise inputs, providing classifications at specific locations. However, performance decreased near the ground and class boundaries due to heterogeneity. To address this, the network was retrained to incorporate spatial information along UAV trajectories, capturing measurement gradients. Various flight patterns, such as scanning and loops, were tested to enhance accuracy. Finally, the method will be applied to UAV measurements over the Lonzée site and validated against LiDAR data. Both numerical and real-data results 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.
Alsteens et al. (Wed,) studied this question.