|
1(89) 2016 MATHEMATICS AND MECHANICS
M.Y. Belikova, S.Y. Krechetova, A.A. Perelygin
Methods and Results of Lightning Discharge Data Clustering
Investigation of spatial distribution of thunderstorms is complex and practically important task. It plays an important part in solving the fundamental problems of atmospheric electricity, as well as lightning protection of buildings and lightning fire protection of forests. One of the data sources of spatial distribution of thunderstorms is WWLLN (World Wide Lightning Location Network). The paper provides the summary of WWLLN data cluster analysis (nearest neighbor algorithm, DBSCAN algorithm). Peculiar features to be taken into consideration are outlined for the data when a certain clustering algorithm is utilized: clusters of different densities, impact of lightning time on clustering results. At the same time, clustering results should be comparable to properties of thunderstorms (for example, thunderstorm mean duration). It is shown that DBSCAN algorithm is capable of processing WWLLN data with peculiarities and thus is the most appropriate algorithm for data clustering. Clustering results can be utilized for further refining and developing of new prediction methods of spatial location and development of thunderstormdynamics.
DOI 10.14258/izvasu(2016)1-17
Key words: cluster analysis, K-means, DBSCAN, WWLLN data
Full text at PDF, 932Kb. Language: Russian. BELIKOVA M.Y.
KRECHETOVA S.Y.
PERELYGIN A.A.
|