Evaluating Future U.S. Tornadic Supercell Development Using the WRF-ARW Model

The United States is home to one of the highest proportions of severe weather on a global scale, with thunderstorms characterized by strong straight line winds, damaging hail, and tornadoes being relatively common, especially in the Central Great Plains.

Climatologically speaking, the Great Plains is perfectly situated for providing a suitable severe weather environment. Warm low-level Gulf moisture from the southeast that contributes to convective available potential energy (CAPE), a dry line between the moist gulf air mass and drier rocky mountain/Chihuahuan desert air masses that provides dynamic lift, and the presence of a veering wind pattern with strong wind shear are common in the region and, when combined, can generate sustained tornadic supercells.

As one may assume, forecasting and evaluating severe thunderstorm development has proven to be challenging, and predicting severe thunderstorm development under future climate scenarios is even more daunting. Capturing fine scale features such as the mesocyclones of supercells requires convective permitting resolutions, and running models over the span of dozens of years at these resolutions would require immense computational power.

Instead of attempting to directly capture individual supercell development in future climate conditions, scientists have treated broader trends in CAPE and wind shear as proxies for severe weather. Diffenbaugh et al. (2013) found that days of decreased wind shear are expected to become more aligned with days of decreased CAPE, while days of increased CAPE will coincide with increased vertical wind shear. This does not offer any indication of change in frequency of severe thunderstorm events but does suggest more intense severe weather outbreaks. Other studies have found that there indeed will be an overall decrease in deep-layer vertical wind shear due to a weakening meridional temperature gradient, but an expected increase in CAPE would outweigh this difference, resulting in an increase in severe weather (Li & Colle, 2013). Overall, there is a growing consensus that when using CAPE and vertical wind shear as primary indicators of severe weather development, there will be an increase in the number of days with favorable environmental conditions for severe thunderstorms (or an increase in strength) within the United States when considering expected changes in greenhouse gas concentrations.

For this project, I am using an alternate method to evaluate severe weather in the future, specifically tornadic outbreaks. A severe weather outbreak that occurred in May of 1999 over the Central Great Plains spawned hundreds of tornadoes and caused significant damages and loss of life. Several EF 3 and 4 tornadoes were generated during the event, as well as an EF 5 that hit Oklahoma City. The purpose of this study is to determine changes in the characteristics of severe tornadic outbreaks due to climate change, and the 1999 outbreak will be used as a representation of such an event. The outbreak has been simulated under historical conditions using the WRF-ARW model and again under expected conditions based on the IPCC’s 5^th Assessment Report (AR5) RCP-8.5 worst case emissions scenario for the 2081-2100 time period.

Each model uses a triple nested domain, with the outermost domain capturing the lower 48 states at a 27-km resolution and the innermost domain ranging from the northwestern corner of Colorado to just off the coast of Louisiana at a 3-km resolution. Six simulations have been run for both the historical and future scenarios to create an ensemble of models that vary in initial conditions, initialized at different times of 6 hourly intervals. The longest running model was initialized at 06z on May 2^nd, and the shortest model was initialized at 12z on May 3^rd. Every model ran until 18z on May 4^th, which covers the duration of the tornado outbreak in the region. To set up the future scenario models the initial and lateral boundary conditions were first calculated for the May averages of the RCP 8.5 models over a 20-year span from 2081-2100 and the historical 1986-2005 CMIP average. The difference between the conditions for each period was calculated, and the difference was then added to the initial and lateral boundary conditions of each historical model to generate the future scenario model initial conditions.

The historical models have proven to be reliable, with model outputs showing rainfall patterns similar to observed TRMM 3-hourly reanalysis data over the same time periods. Models with start times closer to the beginning of the tornado outbreak did appear to have more accurate rainfall patterns.

For evaluation of tornado strength and path length, I am currently evaluating updraft helicity. Multiple studies have found that quantifying updraft helicity at convective permitting resolutions (4 km or less) has proven to be a reliable way of forecasting the number of tornadoes and tornado path lengths (Kain et al., 2010; Clark et al., 2013). Updraft helicity is defined as the local product of vertical wind speed and vorticity integrated over a specified depth, generally 2 to 5 km above ground level. Updraft helicity is computed for both historical and future scenario simulations to evaluate the accuracy of the original model in capturing tornadoes and changes of the outbreak under future scenario conditions.

Few studies have made definitive predictions regarding tornado path length and intensity as a result of climate change, and this study aims to add scientific literature and understanding of the subject. Assuming CAPE and vertical wind shear both increase, I hypothesize there will be an increase in the number, duration, and intensity of tornadoes.