Prediction of cloud-to-ground lightning in the western United States

During the past 30 years, many schemes have been developed to predict lightning (i.e., thunderstorms). These schemes, either extrapolative in nature for the short term or dependent on model output for the longer term, have met with limited success. Yet, more accurate prediction of thunderstorms could help mitigate billions of dollars in annual property damage as well as reduce death, injury, and disruption of human activities. To predict lightning and storms with high flash rates, it is necessary to understand what factors determine when and where thunderstorms develop, as well as determine what factors cause storms to produce high flash rates.; This dissertation focuses specifically on identifying the thermodynamic environment and forcing mechanisms across the western United States that create precipitation systems with lightning. The majority of these convective systems are non-severe. With some thunderstorms, very little precipitation may reach the ground; yet, these “dry” storms spark deadly wildfires in the West every summer.; The goals of this dissertation are: to develop a statistical prediction system that will improve the forecasts of thunderstorms, particularly thunderstorms with high numbers of flashes; to produce forecasts that bridge the gap which exists between extrapolative systems and model-based systems by using both analysis and model forecasts; and to improve the understanding of environmental characteristics which support general thunderstorms and storms with high flash rates.; Predictors are derived from the high-resolution model output (temporal and spatial) of the numerical model known as the Rapid Update Cycle 2 (RUC 2). Additional predictors are from a lightning climatology developed for this study. The RUC 2 model is used in a “Perfect Prog” approach with the predictive equations evaluated using independent data. Using principal component analysis, over 200 candidate predictors from the RUC 2 and the lightning climatology are reduced to a set of 10 new predictors, each representing similar thermodynamic or dynamic processes. Logistic regression is used to produce reliable forecasts of one or more flashes out to three hours.; Comparisons with previous methods show that these forecasts represent a significant improvement in thunderstorm forecasting. Since they have been designed to cover any time period, these forecasts are the first forecasts that fill the gap between current extrapolative techniques and model forecasts covering the critical zero to six-hour time frame. They can be produced quickly from any model analyses or forecasts and are not tied to a specific model. These procedures are also used to successfully predict the probability of convection with higher flash rates and can be easily adapted to predict other lightning related quantities such as positive cloud to ground flashes. Lightning is shown to be especially favored when conditions support a vigorous updraft from the cloud base to at least the −20°C level in the environment. Large numbers of lightning flashes are supported by storms that have vigorous updrafts with higher, colder cloud tops.