Analyzing Doppler LIDAR data to Determine Key Indicators for Vortex Detection

The objective of our research is to create automated methods for analyzing Doppler lidar data to detect and isolate wingtip vortices created by large military aircraft, such as the US Army C17. We aim to use cluster analysis as a tool to classify vortices as separate from the atmospheric background or other features present in the local wind field.

Our algorithms are based on MATLAB's k-means clustering and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) functions. We evaluated the performance of these functions for several different distance metrics. Input to our cluster analysis algorithms was range-height indicator (RHI) scans known to have vortices due to C17 flyovers. The variables included in our analysis were radial wind speed (RWS), signal to noise ratio (SNR), aerosol backscatter (Beta), and the gradient of RWS. To optimize the number of clusters (k) when performing k-means clustering, we performed the cluster analysis multiple times for different values of k and selected the output with the strongest rated clusters based on a metric provided by the MATLAB function. The mean and standard deviation of the variables included in the analysis was calculated for each cluster to determine their identifying features.

After running our algorithms on data known to contain vortices, it was found that clusters can be identified within the vortices whose mean RWS had opposing directions. The wind velocity of one vortex (cluster) would circulate in a more positive direction while the other vortex (cluster) adjacent to it would circulate in a more negative direction. It was also noted that the gradient was the highest in the areas closest to the vortices. Once these factors were identified, our cluster algorithms were applied to the gradient of the variables included in our analysis and we had success locating the vortices more clearly.