Mid-Altitude Radial Convergence Identification Using the KNN Algorithm

Downbursts form and occur rapidly; this makes gathering reliable data on phenomena rare. The k-nearest neighbor algorithm is a machine learning technique that can be effective in applications where there are limited amounts of data. Here we report the implementation of this algorithm in classifying a key meteorological attribute of downbursts - the mid-altitude radial convergence (MARC) of winds. The goal of this project was to provide a proof of concept for the use of the KNN algorithm to successfully differentiate between instances where there was a MARC, where there may have been a MARC, and where there was not a MARC. In partnership with the National Weather Service of El Paso, 17 radar instances of negative, potential, and positive MARC signatures were collected. The parameters used were the magnitude of the difference in speed between the converging winds and the altitude. MATLAB was used to apply the KNN algorithm to the data. The accuracy was evaluated with testing data points. Upon testing the data points, each case returned a correctly identified output indicating 100% classification accuracy. This result is an indication that the KNN algorithm can be used to identify instances where a MARC signature has occurred, however, further testing with more data is required. The results of our research imply that future development is possible to achieve real-time monitoring of radar data to assist in the early identification of MARC signatures, which could allow for life-saving preparation before a downburst.