# Are Tornadoes Getting Stronger?

## Transcription

Are Tornadoes Getting Stronger?

Are Tornadoes Getting Stronger?1 1 Notes for AGU Press Conference. James B. Elsner @hurricanejim December 10, 2013 Our research estimates tornado intensity from the size of the damage path. On average damage paths are getting larger resulting in estimates of increasing tornado strength.2 Thomas H. Jagger & Ian J. Elsner helped with the research. 2 A tornado can destroy an entire town in minutes. Ferocious winds exceeding 220 mph uproot trees and level buildings. It is important to understand if climate change is making them stronger. Problem The EF damage scale used to rate tornado intensity is categorical. We can count the number of tornadoes by EF rating. However, a plot of the number of tornadoes by year and EF rating does not answer the question: Are tornadoes getting stronger? To answer this question we need an estimate of tornado intensity on a continuous scale. How Does a Tornado Form? Strong tornadoes occur from supercell thunderstorms. A supercell thunderstorm is like a regular thunderstorm except the upward moving air is separated from the sinking air. This is because strong winds tilt the top of the thunderstorm relative to its base. If the supercell thunderstorm can survive for at least an hour or so, a tornado is possible. Less than 20% of all supercell thunderstorms produce a tornado. The tornado forms when the air flowing inward gets twisted because winds just above the ground are blowing faster than the winds at the surface. The ribbon of spiraling winds gets lifted into the vertical by upward moving air in the southwest corner of the thunderstorm (see Fig. 2). Additional rotation occurs when the downdraft, that originates at middle levels, wraps around the southwest corner of the thunderstorm. For this reason tornado chasers prefer to position themselves to the south and east of the thunderstorm. Tornadoes occur everywhere in the United States, but are most prevalent east of the Rockies. Increasing awareness and improvements in technology lead to an increasing number of reported tornadoes. Figure 1: Tornado and tornado damage. Figure 2: Illustration of how a tornado forms. are tornadoes getting stronger? 2 Model for Tornado Wind Speed Tornado path length and path width are strongly correlated with EF category. Tornadoes that have longer and wider paths tend to reach higher damage categories. We model this relationship using a categorical logistic regression. Category EF1 N 6179 Length (km) Mean Median 6.24 3.52 EF2 1843 13.05 EF3 543 EF4 120 EF5 13 (1.61,8.05) Width (m) Mean Median 124.2 91.4 (45.7,137.2) 8.85 284.5 20.92 561.3 27.29 809.2 58.95 1389.9 182.9 (3.54,17.03) 26.89 (91.4,365.8) (9.90,35.40) 42.58 (45.51,65.93) 402.3 (228.6,804.7) (16.09,56.73) 67.30 Table 1: Damage path statistics. Data are based on all reported tornadoes in the United States (1994–2011). N is the sample size. The lower and upper quartile values are given in parentheses. Since 1994 the U.S. has been almost completely covered by NOAA’s WSR88D radar. 754.4 (443.5,1063.0) 1307.6 (1207.0,1609.3) Let Wi and Li be the path width and path length of tornado i, then logit[ P( Ti ≤ f )] = (θ f − β 1 Li 1/3 − β 2 Wi 1/2 − β 3 Li 1/3 × Wi 1/2 ) exp(ζ 1 Li 1/3 + ζ 2 Wi 1/2 ) (1) for i = 1, . . . , n and f = 1, . . . , 4, where P( Ti ≤ f ) is the probability of tornado i having an EF category f or lower. 600 400 1 200 0 120 90 60 2 30 Frequency 0 25 20 15 3 10 5 0 10.0 7.5 5.0 4 2.5 0.0 4 3 2 5 We use the square root of the width and cube root of the length so that the model conforms to the proportional odds assumption meaning that the relationship between the EF1 category and all higher categories is the same as the relationship between the EF2 category and all higher categories, and so on. We compute the probability across the EF scale from path length and width then take the product of these probabilities and the set of characteristic wind speeds for each category. The probability in each category is the weight in a weighted average of the characteristic wind speeds. In this way we obtain a single tornado wind speed in m s−1 . We correlate our estimated wind speeds with wind speeds from twelve tornadoes estimated from radar measurements that are independent of the damage. The derived radar wind speeds result from a calibration of mobile radar (Doppler on Wheels) with nearby Weather Service Radar-88D measurements. The wind speeds correlate at .82 (.46, .95) [95% CI]. Although this is a small sample of tornadoes the radar-estimated wind values range from a low of 37.7 m s−1 to a high of 91.2 m s−1 suggesting the 1 0 50 60 70 Predicted Tornado Intensity (m s−1) 80 90 Figure 3: Estimated wind speeds. Figure 4: Doppler on wheels (DOW). are tornadoes getting stronger? potential for our estimates to be broadly applicable throughout the database. Trends Upward trends in path length and width result in upward trends in estimated tornado wind speeds. Upward trends are noted in all EF categories. Increasing effort on tornado surveys over time might result in longer and wider damage paths. This is because the survey typically starts with the most obvious damage and then investigates outward from there, but more research is need to quantify to what extent this has occurred. Figure 5: Trends in path length, width, and intensity. a 2 3 4 5 100 50 2005 2010 2005 2010 2005 2010 2000 1995 2010 2005 2000 1995 2010 2005 2000 1995 2010 2005 2000 1995 2010 2005 2000 0 1995 Annual Mean Path Length (km) 1 b 2 3 4 5 2 1 2000 1995 2010 2005 2000 1995 2010 2005 2000 1995 2010 2005 2000 1995 2010 2005 2000 0 1995 Annual Mean Path Width (km) 1 3 c 2 3 4 5 80 A balance between the pressure gradient force and the centrifugal force produces the fastest tornado wind. The summation of the centrifugal force from the center of the tornado to a radius where the wind blowing the fastest is the mass-specific kinetic energy (E) given by Z E= ro 0 F (r )dr = v2 /2, where v is the fastest wind inside the tornado. The per-tornado kinetic energy by year is shown in Fig. 6. Upward trends are noted at the 50th, 75th and 90th percentile levels. Mass−Specific Kinetic Energy (J/kg) 2000 1995 2010 2005 2000 1995 2010 2005 2000 1995 2010 2005 2000 1995 2010 2005 2000 60 1995 Annual Mean Intensity (m/s) 1 100 4000 3000 2000 1000 1995 2000 2005 2010 Year Figure 6: Per tornado kinetic energy. The quantile trends are shown for the 50th, 75th, and 90th percentile energy. 3 are tornadoes getting stronger? Integrated Kinetic Energy The fastest wind occurs over a relatively small percentage of the total damage path area. The relationship between percent area covered and EF rating is assumed to be logarithmic. Integrating the estimated wind speed over the area covered by that wind speed provides an estimate of the integrated kinetic energy for each tornado. We total the integrated kinetic energy over all tornadoes for each year and then plot the total integrated kinetic energy by year in Fig. 7. A significant upward trend is noted starting at the turn of the century. Integrated Kinetic Energy (GJ) Figure 7: Total integrated kinetic energy. 1.0 0.5 0.0 1980 1990 2000 2010 Year Final Words • Our method and findings suggest an important new domain within hazard research. How does climate change affect tornadoes and other severe local storms? • We have organized the 1st International Summit on Tornadoes and Climate Change to discuss this question with other experts. The summit will be held in Crete, Greece from May 25–30, 2014. • The code to reproduce our results including the intensity model estimates is available at http://rpubs.com/jelsner/TornadoIntensityModel 4