Applying Clustering and Regression Techniques on Lightning and Precipitation Data to Quantify Contributions to the Global Electric Circuit

Abstract:

The global electric circuit (GEC) is the 280kV potential between the Earth and the ionosphere that is thought to be driven mainly by storm systems, including thunderstorms and electrified clouds without lightning. To quantify sources of the GEC, we compare cloud areas from lightning and precipitation rates with fair-weather vertical dc electric field data measured at Vostok, Antarctica (78° S, 107° E) and Barrow, Alaska (71°N, 156°W). The lightning data are from the World Wide Lightning Location Network (WWLLN), and the precipitation product is the Integrated Multi-satellitE Retrievals for GPM (IMERG). In the present work, we develop an analysis tool that applies clustering techniques such as density-based, k-means, and hierarchical methods to the lightning and precipitation data to estimate cloud area every hour. This is accomplished by initially finding large clusters (radius ~ 40 km) that, depending on their density, can be further divided into smaller clusters (radius ~3km). The cluster areas are then found using a convex hull, or, for low-density clusters, a concave hull. The cumulative area of these convex and concave polygons on the globe is the total cloud area. The clustered lightning and precipitation information, along with the electric field data, are then used as inputs for the regression algorithm. The regression analysis output quantifies the contribution of each storm system to the GEC. Our results demonstrate the relevance of our clustering and regression techniques for quantifying how cloud area estimated by lightning, precipitation, or both combined relates to the GEC.