| Author(s) |
|
|---|---|
| Abstract |
The methods used in this analysis included principal components analysis, artificial neural network-based SOM (Self Organizing Map) clustering, cluster validation/evaluation, trend analysis and correlative analysis. The SOM-based classification was applied to datasets of convective available potential energy (CAPE), convective inhibition (CIN), storm relative helicity (SRH), and 2m dew point to create a discrete set of 20 tornado favorable environments (TFEs). Tornado data gathered from the Storm Prediction Center was also mapped to a SOM, to further compare the similarities between the patterns of each of the atmospheric variables with the geographic distribution of tornadoes. After these SOMs were created, a spatial correlation was run on each and compared to the tornado distribution map. A number of different variables would have proven useful for this analysis, herein convective available potential energy, storm relative helicity, 2-meter dew point temperature, and convective inhibition were chosen as ideal for identifying tornado favorable environments. Daily average fields of each of these four variables were acquired from the North American Regional Reanalysis (NARR; Mesinger et al., 2006) project, for the years 1979-2017, at about 123km resolution (every 4th gridpoint of the native 32km resolution), spanning the spatial domain of 30°N to 50°N, and 105°W to 80°W, the area east of the Rocky Mountains, spanning to just before the coast (hereafter referred to as the area of study). Tornado data was retrieved from the Storm Prediction Center archive (SPC, 2019). The necessary data for this research were the date and latitude/longitude coordinate of every tornado touchdown from 1979-2017 in the domain. Tornado data was then binned into counts of tornadoes per day for each 1° x 1° latitude-longitude box across the area of study. The classification produced strong correlations between the spatial distributions of tornado days and atmospheric patterns for CAPE, CIN, dew point, and storm relative helicity (SRH). The results of this study suggest that the patterns with the highest spatial distribution of tornadoes are increasing, at an average of 5.6 more occurrences over this 38-year study period.
|
| Format | |
| Contributor(s) |
Faculty Mentor
Cameron Lee |
| Modified Abstract |
Climate change has become the top concern among scientists as the biggest existential threat to humans. The effects of this change continue to be felt across the globe as extreme droughts, stronger hurricanes, frequent flooding, and other catastrophic environmental disasters become more frequent. The attention given to climate studies such as this can no longer be undervalued. The methods used in this analysis included principal components analysis, artificial neural network-based SOM (Self Organizing Map) clustering, cluster validation/evaluation, trend analysis and correlative analysis. The SOM-based classification was applied to data sets of convective available potential energy (CAPE), convective inhibition (CIN), storm relative helicity (SRH), and 2m dew point to create a discrete set of 20 tornado favorable environments (TFEs).The classification produced strong correlations between the spatial distributions of tornado days and the aforementioned atmospheric patterns. |