Monday, June 10, 2013

The EF-Scale Ratings Brouhaha

In the wake of upgrades to the EF-scale ratings based on mobile Doppler radar-measured windspeeds, there's been a lot of back and forth on the topic of (a) whether or not to use such measurements at all in the rating process, and (b) if mobile Doppler radar measurements are to be used, just how should this be done?

Item (a) is relevant since the NWS Director some time ago issued a directive to the field offices in which use of mobile Doppler radar windspeed measurements should not be used.  The following is a recent directive from the NWS Director, Dr. Louis Uccellini:
Directive NWSI 10-1604, Post-Storm Data Acquisition, requires us to use the EF scale, which is an impact-based rating assigned to a tornado after extensive investigation of the damage it caused. EF ratings are determined by observed damage rather than measured wind because we have no consistent way to measure wind speed for every tornado that occurs. Adhering to NWSI 10-1604 ensures we continue to use consistent methodology throughout the country for assigning EF ratings.

Tornado research is an exciting and rapidly evolving area of science, and we are able to capture more information about the character of tornadoes than ever before. We are updating Directive NWSI 10-1604 to allow the option of including this new information, when available, in the narrative of tornado summaries. The new policy will allow NWS to document available data that are scientifically valid and reliable without changing the objective and consistent EF assessment. Until the update is finalized into policy, WFOs should continue to follow current policy.
This is essentially a management directive from the agency head to the field offices, and it naturally must be obeyed by the NWS employees, without regard to the scientific merits for such a choice.  Personally, I find it absurd to force the offices to ignore the only "direct" measurements of wind speed in a tornado they're ever likely to have.  I'll have more to say about "direct" shortly.  The EF-Scale was never intended to be a damage scale, but rather is a windspeed scale.  The problem always has been that we have so few actual windspeed observations we must use damage to infer the windspeed - otherwise, only a tiny number of events could be rated.  That was Fujita's contribution to the science of tornadoes, and it has been useful (albeit controversial) ever since it was introduced in 1971.

The "consistency" argument is traditional within the NWS as a reaction to technological innovation.  I could provide numerous examples of the wrong-headedness of this policy, but I'll try to keep this as concise as possible.  For instance, should we degrade the data obtained by WSR-88D radars to that from WSR-57s simply to maintain consistency with the older data sets?  Should we disregard the dual polarity information of the new upgrades to the WSR-88D radars just to maintain consistency with the old versions of the radar?  When something new and exciting comes on line, its capabilities should be embraced by the agency, not rejected as inconsistent with older technology!

Turning to item (b), the primary concern is that the mobile Dopplers "measure" the winds within a sample volume well above the standard 10 m anemometer height, and the EF-Scale is supposed to be based on the virtually non-existent anemometer-based windspeeds at that height for a 3-second gust.  By and large, this is a meaningless definition since anemometer measurements in tornadoes are very, very, very rare - a tornado is an anemometer-hostile environment!  [I need to point out that no absolutely "direct" measurement of wind speed is ever possible.  Anemometer output is electronic signals associated with its rotation rate, which must be calibrated to the wind speed.  Any instrument, even in situ systems, are not direct measurements and entail a lot of issues (sensitivity, accuracy, response time, etc.)  Doppler radar windspeed estimates are not absolutely direct, either of course, and are a form of remote sensing, which is why they can be used in tornadoes.]

So the issue becomes:  what relationship exists between the Doppler radar wind measurements and the actual wind at the 10 m level averaged for three seconds [which is virtually never observed but must be inferred from a highly nonlinear relationship with damage]?  As of this moment, research is underway to try to determine this as unambiguously as possible - it will never be completely unambiguous, of course.  There are reasons to believe that windspeeds  might actually increase at decreasing heights as we go downward from where the Doppler measurements are taken.  The details of that windspeed profile remain to be established and there likely is variability from one tornado to another, or even from time to time during the life cycle of a single tornado.  It's unlikely some single profile would actually be observed at all times for every tornado!  Of course, theoretically, the wind must be zero at a height of 0 m but the winds just above that level must increase quite rapidly with height if they are to become capable of damage at 10 m.

Given that the research is not yet complete (and when is research ever truly complete?), it could be argued that it's premature to use the Doppler measurements and the suggestion to keep them but not use them is at least not entirely ridiculous.  However, all the anxiety about the consistency of the EF-Scale ratings strikes me as rather silly.  The existing record is laced with numerous inconsistencies for a host of reasons.  Denying the value of the most direct measurements of windspeed in tornadoes in order to maintain consistency with an inconsistent data set strikes me as silly.  You can argue we shouldn't introduce yet another source of inconsistency but I say we should take advantage of new techology as soon as possible and not get trapped into the "consistency argument" I discussed earlier.

Many years ago, earthquake intensity was "measured" with a damage scale that had many of the same problems as the original F-Scale.  It eventually was replaced with various objective measures of earthquake intensity and has passed into the dustbin of history.  No one ever suggested degrading the Richter Scale to match the old intensity scale for the sake of "consistency" with the older system.  The extent to which Doppler radar measurements will be able eventually to supercede damage estimate is unknown, but it's likely they never will become capable of being used for every tornado to map out the detailed space-time distribution of windspeeds.  Nor will the relationship of their measurements to the mostly hypothetical 3-second gust from an anemometer at 10 m ever be known perfectly.  But to ignore them or defer their use in EF-Scale assessment just for consistency's sake makes absolutely no sense to me.

8 comments:

Mike Dross said...

I think you made some excellent points. I agree, I think not using the mobile Doppler data sets to help determine the peak tornado intensity, which is why the EF/F scale was developed in the first place, is absurd. I think as the science evolves we need to embrace these new data sets, not discount them, I can't imagine scientists in the 1600's not wanting to use the microscope because it invalidated their historical findings. Yes the EF scale is based on damage as an indicator, but that's all we had to base wind speeds on, as a proxy. Other than a tornado that hit an anemometer in Michigan in the 1970's that hit about 150mph before it broke, there has never been a "direct" measurement and probably never will. I say lets embrace the mobile Doppler data and move forward.

Mike Dross said...
This comment has been removed by a blog administrator.
Jonathan Garner said...

Chuck, good blog post. I wanted to make a couple of comments. The use of mobile radar in the assignment of EF scale rating is inconsistent with the methodology used to assign the rating to virtually every other tornado. Different levels of precision are introduced into the F/EF scale database when one tornado is assigned a rating based on radar and all the others are given a rating based on damage indicators. With that being said, I don't deny that the database has been corrupted in various other ways, but that is a different topic. Regardless, in my humble opinion, a consistent methodology is a key to good science. In order for the F/EF scale to be a testable hypothesis, the method used to assign a rating should be applied consistently for each tornado.

George Hrabovsky said...

So, if I am following this—the argument is between using the EF-scale damage manual to infer wind speeds; versus using (even in this case) mobile Doppler radars which infer velocity from the debris and other returns.

The EF-scale is required because it is usually all we have, so there is no choice. We could argue that a slow-moving EF-1 can do EF-2 or EF-3 damage if it is big enough to punch out windows or throw debris through walls and thus over-pressurize the interior causing ceilings and walls to go. Or that an EF-5 can be fast moving enough that it doesn't have the time to completely annihilate everything in its path.

And we are debating this against the prospect of measuring actual droplets or debris caught in the tornado circulation? Admittedly this is not the velocity of the tornado in most cases (there are issues of angular resolution and spectrum width and other deep mysteries to be considered beyond the simple gate-to-gate shear).

And the source of the debate? The wonders of bureaucracy. It is not that one is more accurate than the other, it is that we know more about the inaccuracy of the one than the inaccuracy of the other.

Now, before you laugh with me at the strangeness of this debate. There is a real issue here. We can almost adequately quantity the inaccuracy of the EF scale. We at least understand it better than mystical new technologies that provide us with data that are difficult to interpret. We get a return that tells us that somewhere in a pixel we have a 245 mph wind. What is the error rate? Where in the pixel was the wind speed 245 mph?

Having said that, I think there are similar problems with all methods of verification of tornado winds, and we should be using them all when available. Just because Howie Bluestein isn't going to to up chasing an EF-0 in Rhinelander WI is not sufficient reason not to use in El Reno OK.

Just my two cents worth...

George Hrabovsky said...

So, if I am following this—the argument is between using the EF-scale damage manual to infer wind speeds; versus using (even in this case) mobile Doppler radars which infer velocity from the debris and other returns.

The EF-scale is required because it is usually all we have, so there is no choice. We could argue that a slow-moving EF-1 can do EF-2 or EF-3 damage if it is big enough to punch out windows or throw debris through walls and thus over-pressurize the interior causing ceilings and walls to go. Or that an EF-5 can be fast moving enough that it doesn't have the time to completely annihilate everything in its path.

And we are debating this against the prospect of measuring actual droplets or debris caught in the tornado circulation? Admittedly this is not the velocity of the tornado in most cases (there are issues of angular resolution and spectrum width and other deep mysteries to be considered beyond the simple gate-to-gate shear).

And the source of the debate? The wonders of bureaucracy. It is not that one is more accurate than the other, it is that we know more about the inaccuracy of the one than the inaccuracy of the other.

Now, before you laugh with me at the strangeness of this debate. There is a real issue here. We can almost adequately quantity the inaccuracy of the EF scale. We at least understand it better than mystical new technologies that provide us with data that are difficult to interpret. We get a return that tells us that somewhere in a pixel we have a 245 mph wind. What is the error rate? Where in the pixel was the wind speed 245 mph?

Having said that, I think there are similar problems with all methods of verification of tornado winds, and we should be using them all when available. Just because Howie Bluestein isn't going to to up chasing an EF-0 in Rhinelander WI is not sufficient reason not to use in El Reno OK.

Just my two cents worth...

Anonymous said...

"... Other than a tornado that hit an anemometer in Michigan in the 1970's that hit about 150mph before it broke, there has never been a "direct" measurement ..."

A recent tornado (06OCT2010) moved over a 3-D sonic anemometer and recorded a wind gust of at least 83 m/s.

http://journals.ametsoc.org/doi/abs/10.1175/WAF-D-12-00046.1

The issue of EF ratings based on DI/DOD versus measured winds is discussed in the article.

David Blanchard

Michael Shaffer said...

Very good comments and arguments by Chuck and the named commentors.
If the EF-scale really is a wind speed scale, and not a damage scale, using modern Doppler radars should be a no-brainer.
This, of course, assumes that the algorithms developed by the software engineers, to detect wind speeds for the radars, are actually accurate.
How do we know that the radar algorithms are accurate when we can't measure wind speeds greater than 100 kts or so with physical instruments?
This might be a cart before the horse catch 22 situation.

Chuck Doswell said...

Dave B,

Thanks for your input. Finally read the article ... good job! It's a timely contribution as some of us are working on developing a process to allow the EF-scale to change in an orderly way. It definitely is in need of some revisions.