Spatiotemporal-DBSCAN: A Density-Based Clustering Method for Analyzing Spatiotemporal Ground-lightning Dataset

In recent years, the extraction of extensive spatiotemporal datasets from various sources has surged. One notable example is the cloud-to-ground (CG) lightning dataset in atmospheric area, where each entry records the flash location, including latitude and longitude, and the time the flash occurred.The flashes can be clustered into cells with a set of rules applied to the time-ordered sequence of flash locations, which is called as a thunderstorm cell. Analysis of the cell structure plays an important part in a description of a multicell thunderstorm system in atmospheric area. Clustering algorithms such as K-means and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) cannot be directly applied to these datasets due to their inability to handle spatial and temporal dimensions simultaneously. Our study aims to investigate a clustering method capable of handling large-scale spatiotemporal data in a reliable and efficient manner. The proposed method, called Spatiotemporal-DBSCAN, is applied to both idealized clusters and real datasets, such as the cloud-to-ground (CG) lightning dataset. Experiements demonstrates that the spatiotemporal-DBSCAN performs effectively on spatiotemporal datasets.