Nowcasting of Cloud-to-Ground Lightning with Observations and Model Data

Cloud-to-ground lightning represents a considerable safety threat for aviation operations, particularly at the airport terminals. Baggage-handlers, aircraft refuelers, food caterers, and emergency personnel are all exposed to the risk of being struck by lightning. Lightning forecasts at US airports are typically based upon a combination of data from lightning detection networks and in situ electric field measurements. This presentation addresses the feasibility of using a combination of GOES satellite data, observations, and numerical model data to produce 0-1 hr cloud-to-ground lightning initiation (LI) nowcasts over the Southeastern U.S. and near-shore regions along the Gulf of Mexico.

The initiation of lightning is thought to occur as a result of charge separation in the mixed-phase region of a convective cloud where graupel and supercooled water co-exist, within the so-called charging region of a cumulonimbus cloud. A sufficiently strong updraft is also required to maintain the particle interactions required for charge separation. The University of Alabama in Huntsville's NASA-supported SATellite Convection AnalySis and Tracking (SATCAST) convective initiation nowcasting system has been modified to infer these LI mechanisms from geostationary satellite data. Specifically, SATCAST uses GOES satellite data to obtain proxy fields for locations where the non-inductive charging process is likely occurring. A novel aspect of this study is the use of the GOES 3.9 µm band reflectivity to detect clouds that have recently undergone glaciation. In addition to glaciation, a sharp drop in cloud top temperature detected by the 15-minute trend in the GOES 10.7 µm brightness temperature is also an important indication that the storm cloud is rapidly growing upward, suggesting a very recent and large flux of water and ice in the mixed-phase region of the cloud.

This presentation will investigate the use of satellite-based indicators in conjunction with meteorological observations, trends in existing cloud-to-ground lightning, and forecasts of lightning from models such as NOAA's Localized Aviation MOS Program (LAMP) and the High Resolution Rapid Refresh (HRRR) to nowcast cloud-to-ground LI. These data are combined within several machine learning approaches including logistical regression, random forest and artificial neural networks. Research that uses radar data alone indicates only a ~10-13 min lead-time on LI, whereas this research suggests that a longer lead-time may be made available by using all these datasets together.