Machine Learning-Based Dry Thunderstorm Forecast Model
Project Lead: Wei-Yi Cheng, Department of Atmospheric Sciences
eScience Liaison: Scott Henderson
An example showing the lightning strokes that are located by WWLLN over the U.S. region. Red stars indicate the WWLLN sensors, blue dots indicate the lightning strokes detected by WWLLN.
Lightning is a fascinating phenomenon, but is also a serious threat to human society. Lightning is known to be capable of triggering devastating wildfires in vegetated regions, threatening numerous communities and ecosystems with far-reaching impact in public health and economics. During 2007-2011, U.S. local fire departments estimated an average of 22,600 fires per year that were started by lightning. The lightning-induced wildfires, while happened relatively less frequently, often spread over a much wider area than the ones caused by human, because many of the lightning-induced wildfires occurred in wilderness areas and can spread rapidly unnoticedly.
The lightning-induced wildfires has drawn much attention during the past decades because it is one of the fast-growing natural hazards in most region of the United States. One of the major factors that drives the increase in the lightning-induced wildfire events is because global warming has created a warmer and drier environment in the wildfire-prone area, leading to a higher frequency of “dry thunderstorms”. Dry thunderstorms, traditionally defined as the storms that produce less than 2.5 mm of rainfall, are notable for spreading wildfires for two reasons: (i) they are one of the most common natural origins of wildfires, because there is very little rainfall to prevent the fires from spreading, and (ii) the evaporating precipitation causes excessive cooling of the air beneath the storm, which increases the density of the air as the air descends. When reaches the ground, the evaporatively cooled air spread out horizontally and produce strong gusty surface winds, which help the spread of the fires.
This project aims to improve the dry thunderstorms forecast skill. A supervised ML-based dry thunderstorm forecast model is employed by using the observational lightning data from World-Wide Lightning Location Network (WWLLN) and various atmospheric state variables as features.