Rating Tornado Intensity Based on Moblie Radar

[STexan]

Why can't we have 2 "ratings systems", recorded radar with no/little measurable property damage, and recorded damage with significant impact. Or at the very least put an asterisk next to the radar stated strengths. Future users can utilize the data they choose to. I too think we need both to cover different geographic scenarios, but there definitely needs to be a distinction.

 

[Caleb Chevalier]

The EF5 winds had to be pretty localized. I know there aren't a whole lot of structures south of El Reno, but even with a width of 2.6 miles, this tornado only did up to EF3 damage.
Of course, the El Reno tornado in 2011 was the first one to become upgraded to EF5 by in-situ measurements. And I believe this is the fourth tornado this year to be given an upgrade using the same method.

 

[Jake Orosi]

As some have stated I think that rating tornadoes by mobile radars is a good thing as I have little doubt that it is far more accurate in determining the wind speeds within a tornado than a damage based estimate especially in rural areas where to tornado caused little to no damage. As comes better technology we should use it. I do have issues with it however... if we have for years used a damage based estimate and we now use measured wind speeds from radar on occasion we will likely create a bias in tornado strength climatology, especially so as mobile radar ratings become more frequent. 20 years from now when we are reviewing tornado strength since records began, 2013 may show up as a year that starts a spike in more intense tornadoes solely because we have rated so many by mobile radars this year. We'll be looking back wishing we could make some more accurate comparisons between damage rated tornadoes and mobile radar rated tornadoes.

I'm not suggesting that we not use these radars for rating but rather I suggest that we also note the damage survey results as well in our climatological records for comparisons sake. Using a satellite remote sensing analogy (I'm a GIS guy), when new technologies with better satellite resolutions come along it is certainly used but the older satellite with less resolution is still used so that the two satellites can be calibrated to each other. With this calibration satellite images from 2013 can be directly compared to satellite images from 1983 despite the change in resolution (with resolution: the accuracy of the the data). I suggest that the NWS/SPC/NCDC also record the damage indicated ratings as well for a fair comparisons sake. Thus the Rozel, KS tornado would be noted as an EF-2 by damage, EF-4 by mobile radar... similarly the El Reno tornado would be noted with both an EF-3 and EF-5 ratings (along with other tornadoes rated by mobile radar). Maybe a new scale other than EF should be used for a rating based on mobile radars?

These are just my humble opinions and I hope that we do find some way of keeping or noting these rating scale differences in tornado climatology records for sake of future research. I understand the EF-Scale has flaws but no matter the flaws we need a way of comparing a older more flawed rating to a newer less flawed rating.

If there is that much of a discrepancy between damage-derived estimated wind speed and instrument-measured wind speed, all that means is that our earlier damage-derived estimates were inaccurate. It shouldn't be any great matter to correct this situation: we need at most around ten years of concurrent but independent measurements of both damage and instrument readings - within the same storms as much as practicable. By that time, we should be able to establish a consistent and predictable relationship between EF-ratings derived from damage and measured wind speeds of those storms. That will let us create a formula that we can apply backwards to historical tornado data to modernize it.

Maybe at that time a new scale can be created - the Modern Fujita; or the current one can be retained and the wind speed values tweaked - perhaps in the future all it'll turn out we need to do is add an EF-6 and -7.

I know one thing for sure: it's the opposite of progress to sacrifice or disregard accuracy for the sake of the appearance of consistency.

 

[Jeff Duda]

Why can't we have 2 "ratings systems", recorded radar with no/little measurable property damage, and recorded damage with significant impact. Or at the very least put an asterisk next to the radar stated strengths. Future users can utilize the data they choose to. I too think we need both to cover different geographic scenarios, but there definitely needs to be a distinction.

Or we could keep things simple and go with what we have. I'm not sure why so many of the forum members think it is so bad to use Doppler radar measurements to rate tornadoes. If you think it is compromising the tornado database, realize how many inaccuracies, inconsistencies, and missed ratings there already are in the database due to poorer technology/reporting in prior years. Radar and in-situ measurements are a step in the right direction - they give actual wind speed measurements rather than estimates based on damage. That was the goal of the Enhanced Fujita scale all along.

 

[STexan]

I have to question the accuracy of wind speed measurement in a meso with very heavy rain and hail densities obscuring radar penetration. I may be way off and maybe radar may have no problem penetrating AT GROUND LEVEL some mesos with heavy and wide ranging rain/hail equivalent to excess of 3-4 inch/hr. I will defer to others on this aspect. If the measurements can be 100% relied upon for real-world accuracy of speed of rain and debris, I would tend to agree with the idea perhaps we should ONLY use radar based assessments in the future. How accurate can radar speed measurements be at the fringe of the range say on smaller tornadoes that may also even have a very short life span? Is Dual Pol extensive enough yet?

 

[Jake Orosi]

Maybe - but maybe not. Maybe a greater number of particulates like rain and hail might make radar readings more accurate, because their motion is subject to the wind and so could brighten its reflectivity. Or maybe not. We need more data!

Seems there's still some actual tornado science left to do after all.

 

[Jeff Duda]

I have to question the accuracy of wind speed measurement in a meso with very heavy rain and hail densities obscuring radar penetration. I may be way off and maybe radar may have no problem penetrating AT GROUND LEVEL some mesos with heavy and wide ranging rain/hail equivalent to excess of 3-4 inch/hr. I will defer to others on this aspect. If the measurements can be 100% relied upon for real-world accuracy of speed of rain and debris, I would tend to agree with the idea perhaps we should ONLY use radar based assessments in the future. How accurate can radar speed measurements be at the fringe of the range say on smaller tornadoes that may also even have a very short life span? Is Dual Pol extensive enough yet?

Hopefully Jeff Snyder comes on here eventually and answers this question himself, because he has a better grasp on the theory than me. The radar they used is X-band, so it does suffer from attenuation in heavy rain and hail. However, they mitigate that by being very close to the action. Attenuation should not severely hamper measurements of a rain-wrapped tornado sampled at 4-5 km range with X-band. Yes, there is error associated with the electronic system (it's stochastic and should have a very low bias) as well as with the spectrum width (which he has already said was very high), but that just means the wind measurement is more like 295 mph with an uncertainty range of perhaps 20 to 30 mph in either direction. I doubt the uncertainty is higher than 100 mph.

 

[MClarkson]

I have to question the accuracy of wind speed measurement in a meso with very heavy rain and hail densities obscuring radar penetration.

The Doppler shift will be the same whether its a 296mph drizzle drop or a 296mph hailstone. Total energy returned to the radar(reflectivity) would change based on target size, the frequency change would not. While eventually, with increasing range, you get so much crud that that you simply get no signal there and back and don't get any measurement at all, if you are getting returns I think they are quite accurate. A bigger problem I think would be a sample off the surface, but its not that far off the surface. A boundary layer specialist might be able to answer that more precisely than me.

 

[JamesCaruso]

Tornadoes are typically EF-rated based on damage, true; but remember the whole point of that: the extent of the damage is used as a gauge to estimate wind speeds which are usually extremely local and therefore not measured by actual instruments that were not in the path of the tornado. Certain wind speeds are expected to cause a certain severity of damage. It's exactly like using the Beaufort scale when you don't have an anemometer.

.

Jake wrote what I was thinking...

The purpose of the F / EF scales was to use damage to estimate wind speeds that could not otherwise be directly measured at the time they were occurring. Now that we are capable of directly measuring the winds of a tornado using mobile radar, we absolutely should use those direct measurements in the ratings.

As far as inconsistencies in the database, as others have noted there are already plenty of those for numerous reasons. Isn't it also inconsistent to rate two 200 mph tornados differently just because one hit a town and one stayed over open land?

 

[John Peters]

Here is something to think about:

Many tropical cyclones prior to the era of hurricane hunter aircraft and satellite imagery have Saffir Simpson ratings. The majority of these ratings are likely from "fortuitous " in situ observations.

Now we have hurricane hunter aircraft obs for storms that get sufficiently close to land, and satellite derived intensity measurements. Should we throw these out, because they introduce inconsistencies to the data base? Most of you would consider that laughable.

I think the situation I have described is at least somewhat analogous to what is being discussed in this threat.

 

[Jeff Snyder]

Hopefully Jeff Snyder comes on here eventually and answers this question himself, because he has a better grasp on the theory than me. The radar they used is X-band, so it does suffer from attenuation in heavy rain and hail. However, they mitigate that by being very close to the action. Attenuation should not severely hamper measurements of a rain-wrapped tornado sampled at 4-5 km range with X-band. Yes, there is error associated with the electronic system (it's stochastic and should have a very low bias) as well as with the spectrum width (which he has already said was very high), but that just means the wind measurement is more like 295 mph with an uncertainty range of perhaps 20 to 30 mph in either direction. I doubt the uncertainty is higher than 100 mph.

Attenuation is appreciable at X band -- we've seen signal extinction is as little as 8-10 km into a storm. In this particular case, our deployments were NE and E of the tornado with relatively little intervening precipitation. As long as SNR is high, there's still plenty of room to retrieve good phase estimates (and thus good velocity estimates) even in the presence of attenuation. See Snyder et al. (2010) in JTECH for a description of some ways we try to mitigate and "correct" for attenuation of our mobile radar data.

It's important to remember what it means when we say a particular gate has XX m/s radial velocity. The calculated radial velocity is a reflectivity-weighted mean velocity - it's the radial component of the speed of the scatterers within the resolution volume (weighted towards the center of the beam on account of the shape of the beam sent out using a parabolic antenna/dish). Of course, not all scatterers are moving at the same speed and direction, so what's being "measured" is actually some weighted combination of everything. The most unlikely thing to actually be measured is the wind (since air molecules have too small of a backscattering cross-section to be measured by Xband weather radars)! Hydrometeors tend to respond the quickest to changes in wind speed and direction, but even they have a lag that results in a mis-match between hydrometeor motion and air motion. If you consider higher inertia items like large debris, especially in areas of extreme wind accelerations, this can really get interesting...

The number that's currently public and that we have very confidence in is 132-133 m/s. These extreme winds were associated with subvortices or perturbations along the south side of the tornado. Given our location, the radial velocities measured are the sum of the translation of the tornado towards the radar + the translation of the subvortices (which was unbelievably fast!) + the flow around the subvortices. Geometrically, we were in a good position to nearly maximize all three of these components. These data are still preliminary, and I'm still going through what we've collected. I would expect these numbers to change as we investigate further.

I spent a lot of time in the fields where the highest winds were observed, and they literally are open wheat and vegetation (weeds?) fields. The DIs that we examined matched up relatively well with the radar measurements, which, combined with the massive "excess" of speed over the eF4/EF5 threshold, leads me to be confident that winds in the EF5 range were experienced at 2-10 m AGL. I'm running the math know, and the standard deviation expected from the observations is considerably less than 20 mph.

 

[Shawn Schuman]

I spent a lot of time in the fields where the highest winds were observed, and they literally are open wheat and vegetation (weeds?) fields. The DIs that we examined matched up relatively well with the radar measurements, which, combined with the massive "excess" of speed over the eF4/EF5 threshold, leads me to be confident that winds in the EF5 range were experienced at 2-10 m AGL. I'm running the math know, and the standard deviation expected from the observations is considerably less than 20 mph.

Thanks for weighing in Jeff. This was just east of 81, right? I was trying to find photos of those areas today to see whether there was anything of note, but I haven't found much yet. Of course, I doubt most people are very interested in taking photos of empty fields anyway. Poking around on Google Earth it doesn't look like there were even very many trees in that area.

 

[Dan Robinson]

I crisscrossed the damage path the next day looking for my camera's lens hood, which the RFD took from me. There is very little that the tornado could hit in that location, as others have said. Houses were widely scattered, most a mile apart. We know that most EF5/F5 tornadoes produce that level of damage in only a very tiny portion of the total path, sometimes only one structure gets the tornado a 5 rating. It's no surprise then that EF5 DIs would be scarce or non-existent in this instance.

The tornado's size and violent appearance in person certainly indicated a high-end event, and I expected at least EF4. I'm curious to hear if any ground scouring will be found. The gravel roads seemed to be intact, that is I didn't see any segments where substantial aggregate material had been removed. Granted, I hadn't covered all of the road grid in the vicinity.

 

[Jeff Snyder]

Thanks for weighing in Jeff. This was just east of 81, right? I was trying to find photos of those areas today to see whether there was anything of note, but I haven't found much yet. Of course, I doubt most people are very interested in taking photos of empty fields anyway. Poking around on Google Earth it doesn't look like there were even very many trees in that area.

Here's a picture I took near 10th and Radio looking to the northeast:

Much of the area looks like this. The picture above is of a field of some very stalky plant/weed (it looks "unfarmed"); other areas are wheat. As you note, there's almost nothing to hit.

Regarding scouring... I'm not sure what's been found. I saw a lot of matted down wheat, but I'm not sure what that tells us. As far as I know, there is almost no experimental data on the relationship between ground scouring and wind speeds. I suspect that scouring is highly sensitive to debris loading (which there probably wasn't much of given fields like the one above), residence time of the highest winds over the land, soil conditions (moisture, etc.), plant conditions (root depths, etc.), and maybe even wind variance (e.g., did the high winds start immediately or did they ramp up?). I know there's anecdotal evidence for scouring by high-end tornadoes in the past, but the lack of experimental data makes me very nervous to use that as a non-official DI. I did see areas in which the vegetation was scalped off 2-3 inches above ground (or even to ground level in at least one location), but those areas were "uptrajectory" of debris (which is what makes me think that debris loading has some effect on the appearance of what's been termed "ground scouring" in the past).

 

[Eric Edgar]

Should those in the scientific community consider starting a new rating scale, to be used PARALLEL to the current Enhanced Fujita scale? For example, the recent El Reno event would be rated EF3/R5 - EF3 for "ground truth" damage assessment, R5 for radar-indicated (slightly above ground) wind speed. This would allow continued comparison across the database (both with events prior to the advent of mobile radar, and with new events that do not get sampled by mobile radar), while not "throwing away" more accurate data from the mobile radar.

It just seems that we need to be as explicit as possible about why an event gets rated the way it does, otherwise we are introducing more and more variability into the dataset. Taking the logic a step further, mobile radar is going to continue to evolve - at some time in the future it may be more precise than it is now, and it may be more capable of measuring ground-level winds than it is now (I don't understand the technology so I don't know how much more precise/accurate it can get).

I suppose asterisks in radar-derived situations would achieve the same result.