Data-Driven Classification of Hydrometeor and Aeroecological Observations of Polarised X-band Doppler Radar

Speaker: Maryna Lukach

Maryna Lukach, Ryan R. III Neely, Jonathan Crosier, David Dufton, Lindsay Bennett, Christopher Hassall, Elizabeth Duncan, William Kunin and William Evans

A correct, timely and meaningful interpretation of polarimetric weather radar observations requires a well-developed technique that automatizes recognition of separate classes. As observed polarimetric variables represent nonlinear processes, any application of simple thresholding may lead to wrong classification of observed data and clustering technics can help to solve this problem. Various clustering approaches developed over the last decades provide wide range of tools for classification of multivariate data of any origin and nature.

This webinar presents a novel data-driven technique for identifying different classes in Quasi-Vertical Profiles (QVPs) based on observations made by the NCAS X-band dual-polarization Doppler weather radar (NXPol). This technique provides a physical delineation of multivariate radar observations into classes. The classes are identified as clusters belonging to a hierarchical structure preserving the data-driven inheritances. The number of different classes in the data is not predefined and the method obtains the optimal number of clusters by implementing a recursive process. The obtained optimal clustering is then used to label the original data.

This technique is applied to the identification of hydrometeor types and their associated microphysical processes and to the detection and identification of aeroecological classes in the QVPs. Although this demonstration uses NXPol data, the technique is generally applicable to similar multivariate data from other radar observations. 

Seminar20200713Lukach.mp4

Wednesday July 15, 2020, 13:00

The seminar will be broadcast live via Zoom, please follow the link below.

https://zoom.us/my/ncasseminarseries 

Password: 619447  

Please note these seminars are intended to be internal to NCAS.  Please do not share details of the link without prior permission.

Please follow these simple guidelines to help make our seminar successful.