jpeters3 Posted June 3, 2011 Share Posted June 3, 2011 It certainly is statistically robust. A climatology of actual radial wind measurements in 70 tornadoes, 50+ of them with obs within 500 m of the ground is, in fact, the best evidence we've ever had to evaluate the EF scale. The sampling of strong to violent tornadoes is very poor - in fact, there is only one EF-5 rated tornado within the data. If this same analysis were conducted on say, 1000 tornadoes, we may find that the number of EF-4/5 rated tornadoes matches, percentage wise, much better with those determined to be EF-4/5 by DOW measurements. I'm not speculating as to whether or not this would be the case, but the small sample size for violent tornadoes is far less than sufficient for statistically robust conclusions. Link to comment Share on other sites More sharing options...
thewxmann Posted June 3, 2011 Author Share Posted June 3, 2011 Perhaps I should've worded by title a bit better - I wrote in my blog post that I had redefined violent, in terms of the study, for EF3 and over, because the sample size of EF4/5 tornadoes was too small for a good analysis. Regarding the path length thing, it is also biased since in general, 30 EF3+ tornadoes in March/April will have a longer path length than 30 EF3+ tornadoes in May/June since the jet stream is stronger in the early season. Time on ground, I believe, would be the most unbiased measurement - but only start/end time is reported in the tornado database (I would have to write a program to calculate the path time in bulk). At any rate, I normally include path length in my analysis of tornado statistics, but for this case I wanted to do it quickly and didn't have the time to go through and methodically find the path lengths of all the EF3+ tornadoes this year. Link to comment Share on other sites More sharing options...
jpeters3 Posted June 3, 2011 Share Posted June 3, 2011 I think some of them definitely do and the radar climatology shows the preferred strength of tornadoes is EF-2 rather than EF-0 I agree that the length of time a tornado is strong or violent has a lot to do with its threat to life and property. However, a surveyed tornado suffers from the same weakness as the radar study in that a tornado is rated at its highest damage indicator. Most violent tornadoes are actually only violent for brief periods of time if you go by the damage surveys (probably because there's usually not long paths of structures), so it's difficult to know what fraction of the time violent tornadoes are actually violent. So I'm not sure that it really helps us that much, though I see what you're saying. I think we'd need to know more about the relative frequency of violent winds in the violent tornadoes. Unfortunately long tornado path lengths mean long distances where there's just not very much around to adequately evaluate the wind speeds. I think the most interesting result out of this paper is what we both touched on - the fact that most tornadoes probably do achieve at least EF-2 winds at something through their life cycle, and I think this points to an inadequacy toward the lower end of the scale. It does suggest that there is also an inadequacy toward the upper end of things (especially since 4-5 tornadoes were EF-5s per DOW, and only one was rated EF-5 per damage), but I still think more data is needed. I think I'm also starting to more comprehensively understand your argument regarding the EF scale as well - I think the solution (for peace of mind) might be to think of an EF scale rating in more probabilistic sense... i.e. based on the available data, an EF-4 rated tornado most likely had wind speeds in this range (with the knowledge that we will likely never know the true maximum winds and how long they were attained for). With all that being said, I still think there is plenty of subjective evidence that tornadoes of 2011 have been more intense, or intense for more sustained intervals based on some of the damage surveys. Hackelburg, for instance, produced numerous instances of EF-5 damage along a 100+ mile path, while, say, yazoo city, only produced a few isolated instances of EF-4 damage. Tuscaloosa-Birmingham produced high-end EF-4 damage in both Tuscaloosa and Birmingham. It appears as if there were numerous instances of incredible damage along the track of Piedmont-El Reno. The joplin tornado produced > 1 mile sustained swath of EF-5 damage through Joplin. Yes, we have seen previous tornadoes produced sustained violent damage along their paths, but usually once a year - if not once every few years. This year there have been many. Link to comment Share on other sites More sharing options...
AtticaFanatica Posted June 3, 2011 Share Posted June 3, 2011 The sampling of strong to violent tornadoes is very poor - in fact, there is only one EF-5 rated tornado within the data. If this same analysis were conducted on say, 1000 tornadoes, we may find that the number of EF-4/5 rated tornadoes matches, percentage wise, much better with those determined to be EF-4/5 by DOW measurements. I'm not speculating as to whether or not this would be the case, but the small sample size for violent tornadoes is far less than sufficient for statistically robust conclusions. Huh? There's only one damage-rated EF-5 tornado because the damage ratings are wrong, there were actually 11 they observed (10 of them were rated wrong), that's a pretty good sample. It's not their fault that the survey ratings of those tornadoes are poor. Link to comment Share on other sites More sharing options...
jpeters3 Posted June 3, 2011 Share Posted June 3, 2011 Huh? There's only one damage-rated EF-5 tornado because the damage ratings are wrong, there were actually 11 they observed (10 of them were rated wrong), that's a pretty good sample. It's not their fault that the survey ratings of those tornadoes are poor. I'm not saying the study was done poorly... The comparison is between the number of violent rated tornadoes and the number of dow indicated violent tornadoes. All that I'm saying is that the number that were addressed in this data set falls far short of what you'd need for statistically convincing results. Take the EF-5 ratings for instance - there are 11 candidates for EF-5 in this study... the fact that only one was rated EF-5 could be mere happenstance, and if 11 additional EF-5 candidates were analyzed, there may be better correspondence. A larger sample (much larger) would be necessary to ensure statistical significance. So to make a long story short, the results with respect to violent tornadoes are suggestive, but they are not statistically significant Link to comment Share on other sites More sharing options...
jpeters3 Posted June 3, 2011 Share Posted June 3, 2011 The reduction in wind speeds at the surface, based on in situ tornado measurements, is small. How many instances do we have in situ measurements and corresponding DOW readings at elevation for comparison? I'm not suggesting that you are wrong, I'm just curious. Link to comment Share on other sites More sharing options...
AtticaFanatica Posted June 3, 2011 Share Posted June 3, 2011 I think the most interesting result out of this paper is what we both touched on - the fact that most tornadoes probably do achieve at least EF-2 winds at something through their life cycle, and I think this points to an inadequacy toward the lower end of the scale. It does suggest that there is also an inadequacy toward the upper end of things (especially since 4-5 tornadoes were EF-5s per DOW, and only one was rated EF-5 per damage), but I still think more data is needed. I think I'm also starting to more comprehensively understand your argument regarding the EF scale as well - I think the solution (for peace of mind) might be to think of an EF scale rating in more probabilistic sense... i.e. based on the available data, an EF-4 rated tornado most likely had wind speeds in this range (with the knowledge that we will likely never know the true maximum winds and how long they were attained for). With all that being said, I still think there is plenty of subjective evidence that tornadoes of 2011 have been more intense, or intense for more sustained intervals based on some of the damage surveys. Hackelburg, for instance, produced numerous instances of EF-5 damage along a 100+ mile path, while, say, yazoo city, only produced a few isolated instances of EF-4 damage. Tuscaloosa-Birmingham produced high-end EF-4 damage in both Tuscaloosa and Birmingham. It appears as if there were numerous instances of incredible damage along the track of Piedmont-El Reno. The joplin tornado produced > 1 mile sustained swath of EF-5 damage through Joplin. Yes, we have seen previous tornadoes produced sustained violent damage along their paths, but usually once a year - if not once every few years. This year there have been many. These are all fair points. Link to comment Share on other sites More sharing options...
AtticaFanatica Posted June 3, 2011 Share Posted June 3, 2011 I'm not saying the study was done poorly... The comparison is between the number of violent rated tornadoes and the number of dow indicated violent tornadoes. All that I'm saying is that the number that were addressed in this data set falls far short of what you'd need for statistically convincing results. Take the EF-5 ratings for instance - there are 11 candidates for EF-5 in this study... the fact that only one was rated EF-5 could be mere happenstance, and if 11 additional EF-5 candidates were analyzed, there may be better correspondence. A larger sample (much larger) would be necessary to ensure statistical significance. So to make a long story short, the results with respect to violent tornadoes are suggestive, but they are not statistically significant I'm sure the peer-reviewed article will touch on significance whenever it comes out (with even more data I assume since it's been another 3 years), but I think the idea is that all the tornadoes as a whole constitute a sizable sample, and it can be said as a whole that the ratings are inaccurate. I see what you're saying about EF-5's in particular, but man, 1 for 11 is pretty bad. Link to comment Share on other sites More sharing options...
AtticaFanatica Posted June 3, 2011 Share Posted June 3, 2011 How many instances do we have in situ measurements and corresponding DOW readings at elevation for comparison? I'm not suggesting that you are wrong, I'm just curious. I'd estimate <5 from the TIV, the CSWR tornado pods, and the TWISTEX HITPR probes, just a guess though. Link to comment Share on other sites More sharing options...
jpeters3 Posted June 3, 2011 Share Posted June 3, 2011 It would bee interesting to catalog a large sample of gate-to-gate vortex sigs from WSR-88D network, and try to compare this to the EF scale ratings - this is mentioned on the article, along with the fact that EF-scale ratings and DOW readings weren't really well correlated with stationary radar velocity data, and I"m not surprised given the variable sampling heights and potential discrepancies between tornadic circulation at considerable height and near the surface... I wonder how big of a sample they looked at? Link to comment Share on other sites More sharing options...
AtticaFanatica Posted June 4, 2011 Share Posted June 4, 2011 It would bee interesting to catalog a large sample of gate-to-gate vortex sigs from WSR-88D network, and try to compare this to the EF scale ratings - this is mentioned on the article, along with the fact that EF-scale ratings and DOW readings weren't really well correlated with stationary radar velocity data, and I"m not surprised given the variable sampling heights and potential discrepancies between tornadic circulation at considerable height and near the surface... I wonder how big of a sample they looked at? 88D data are probably too high up off the surface to be useful. Even if you only used a sample of tornadoes within 50, or 25 or 10 km of the 88D network, the spatial resolution is still relatively large, so the size of the radar's resolution volume precludes very accurate max wind readings. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.