SODAR Echograms Based Model Development for Atmospheric Boundary Layer Characterization

Abstract

SOnic Detection And Ranging (SODAR) is a well-known and widely accepted meteorological tool for supplying continuous real-time and critical Atmospheric Boundary Layer (ABL) data. Data is critical for evaluating environmental impact assessments and city-specific carrying capacity for pollutants. Existing SODAR technology was improved, which included, acoustic antenna advancements, virtual instrumentation, and improved data processing approaches. This advancement will affect the observed data, and data will be more accurate as a result of calibration and testing of equipment and materials. An acoustic antenna was designed using moving-coil transducers, parabolic dish, and acoustic baffle. Several types of Aluminium Composite Panel (ACP) for acoustic baffle were tested to their characteristics like Sound Transmission Coefficient (STC) and Noise Reduction Coefficient (NRC) in the reverberation chamber. A comparison investigation was carried out on transmission loss and absorption. It was concluded that baffle (ACP with foam) is the suitable material with STC (34) and NRC (0.98) for an acoustic antenna. The SODAR echogram for the ABL structure was derived and successfully applied in a highly accurate and reliable machine-learning method. In terms of performance, five functional selection procedures and eight classification methods were examined. From 1698 SODAR echograms, 133 statistic features were calculated. Machine-learning methods were used to ensure the unbiased estimation of different structures. Ten cross-validations were used to determine accuracy. The boosted tree classifier was given the strongest prognostic presentation with 133 features (total prediction rating of 52.02%). After applying the Laplacian method for feature selection, the classifier (overall prediction performance 62.19%) showed the highest prognostic presentation with 20 features. The large variability analysis indicates the choice of a classification method for performance variation.