A dynamically downscaled climatology of severe convective thunderstorms over the United States
Abstract
Severe convective storms—and the tornadoes, hail, and damaging winds that they produce—represent a real risk to life and property in the United States. In 2011, tornadoes claimed over 540 lives over 15 different states, the fourth highest toll on record. It is not known whether 2011 represents a positive trend in the frequency of severe thunderstorms and associated phenomena because of the many uncertainties in the historical record. This study seeks to examine a method developed to study these severe phenomena using a high-resolution numerical weather prediction model driven by large-scale reanalysis and employing an artificial neural network. Specifically, high resolution simulations of the Weather Research and Forecasting model, forced by data from the NCEP/NCAR Reanalysis Project are used to generate daily re-forecasts of every day in April, May, and June over 1990–2009. These simulations are examined at hourly intervals for the presence of severe convection through the use of an artificial neural network that was trained using days from ten years of modeled data that coincided with observed days of severe weather. This network is applied over all twenty years of data in order to examine the modeled trends in severe phenomena. Tests are also performed to examine the sensitivity of the modeling procedure to various parameters including integration time, land surface heterogeneity, and model physics. Results indicate that, contrary to the biased observational record, there has been no statistically significant change in the frequency of warm season severe weather occurrences in the eastern two-thirds of the United States over 1990–2009. Further, sensitivity tests indicate an approach with more frequent re-initializations of the model produces a more physically realistic climatology of severe convective events, and that this climatology is fairly insensitive to small changes in the land surface and model physics. A similar result is found for the spatial distribution of heavy and extreme rainfall events.
Degree
Ph.D.
Advisors
Trapp, Purdue University.
Subject Area
Atmospheric sciences
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.