Investigating Atmospheric Discharges at the Tops of Thunderclouds with Machine Learning

Research output: Book/ReportPh.D. thesis

38 Downloads (Pure)

Abstract

Blue Corona Discharges (BCDs) are streamer discharges characterized by ionized plasma propagating as an ionisation wave. Their profuse activity was first discovered by an astronaut on the International Space Station (ISS) before being observed by the Atmosphere-Space Interactions Monitor (ASIM). In the photometers of ASIM they present with an optical signature in the second positive band of Nitrogen (N22P) at 337 nm (blue) but with no corresponding emissions in the OI line at 777.4 nm (red) which is usually associated with hotter discharges. Typically they emanate from the tops of thunderclouds and they are correlated with more severe weather than lightning flashes are. This thesis investigates both the optical properties of BCD events and how they relate to various important convective and meteorological parameters. In the first paper, we find that the events can be reasonably separated into fast and slow events depending on their originating cloud depth with an average cloud depth difference between fast and slow of around 1 km. Though speculative, this distance could be indirectly measuring the charge layers of the thundercloud. Another atmospheric discharge of interest in this work are Terrestrial Gamma-Ray Flashes (TGFs) which are high energy flashes of X- and gamma-rays believed to be caused by runaway electron avalanches that lose energy through bremsstrahlung. TGFs can also be observed by ASIM and understanding their correlation with meteorological conditions is valuable, both to understand their production mechanisms but also to understand the limitations of detecting TGF events from Low Earth Orbit (LEO) observatories.

In the second paper we show that while TGFs tend to favour lower Cloud Top Temperatures (CTT) as has been shown in other works, they do not favour more severe weather. The proportion of events coming from systems with overshooting tops s similar between TGF producing systems and systems that produce only lightning, as is the Convective Available Potential Energy (CAPE). We suggest that the reason for TGF systems favouring lower CTT systems when observed by LEO observatories is related mainly to the attenuation of the gamma-rays from lower altitude TGFs, which reduce their intensity to below the detection threshold of the observing LEO instrument. As such, weaker, lower altitude TGFs should be detected if observed from closer to the cloud from, for instance, a plane. Finally, we use the information gained from the first two papers to construct a predictive model for the BCD events in the third paper. In order to allow us to see changes in the predicted production of these events we base the model on purely reanalysis data which lets us run the model for 1980-2022. During this time period, the model suggests an increase of about (0.13 ± 0.04)% per year. As a function of temperature it predicts an increase of (7 ± 2)%K−1. These increases are seen mainly in central Africa, Australia and eastern Europe. We suggest that the correlation of BCDs with more severe weather than lightning may explain why the model predicts an increase of BCD events while regular lightning flashes may be constant or even decrease. Most lightning prediction models suggest that while the number of lightning days may decrease, the severity of the remaining lightning days increases significantly. This model is a step towards constructing global models for the prediction of atmospheric discharges, lightning in particular, as currently regional models disagree on the magnitude and even sign of the change in lightning production in a warming climate.
Original languageEnglish
Place of PublicationKgs. Lyngby
PublisherTechnical University of Denmark
Number of pages96
Publication statusPublished - 2024

Fingerprint

Dive into the research topics of 'Investigating Atmospheric Discharges at the Tops of Thunderclouds with Machine Learning'. Together they form a unique fingerprint.

Cite this