Paul McCrone




ATS 615

RADAR and Severe Storms




Prof: Dr. Richard Penc

Date:  December 11, 1998




Creighton University

Dept. of Atmospheric Sciences

Omaha NE



Figure 1: Developed by the Author.


Figures 2 through 7: Taken from Rotunno, R., 1993: Supercell Thunderstorm Modeling and Theory. Geophysics Monograph 79, The Tornado: Its Structure, Dynamics, Prediction, and Hazards , 57-73.


Figures 8 and 9: Finley, C. 1990:  Technical Training: Doppler Weather Radar: Thunderstorm Morphology and Dynamics, 27 August 1990,  Chanute Technical Training Center (ATC), 3350 Technical Training Group, Chanute Air Force Base, Illinois.


Figure 10: Taken from the National Severe Storms Laboratory (NSSL) Web Page,


Figure 11: Taken from Lemon, L. R.,  D. W. Burgess, and R. A Brown, 1978: Tornadic Storm Airflow and Morphology Derived from Single-Doppler Radar Measurements. Mon. Wea. Rev., 106, 48-61.


Figures 12 and 13: Taken from the National Severe Storms Laboratory (NSSL) Web Page,


Figures 14 through 20: Taken from (Author Unknown) 1998: WSR-88D Build 10 Training, National Weather Service, NEXRAD Operational Support Facility, Norman Oklahoma, available at









The focus of this paper will be a discussion of the current state of the art related to 


the detection of mesoscale cyclones (or ‘mesocyclones’) using Doppler Weather


RADAR.  A brief discussion of the theory on mesocyclone development will be followed


by some specific applications, which will focus on the WSR-88D (Next Generation


Weather RADAR – or NEXRAD). Particular attention will be paid to the use of


computer based algorithms in the automated detection of mesocyclones, in addition to a


discussion of pattern recognition in both reflectivity and radial velocity fields. In addition


to this background, we will look at two different scenarios that demonstrate the detection


of  mesocyclones.







Mesoscale cyclones (or ‘mesocyclones’) are indeed significant meteorological


phenomena to study.  They often precede tornadoes, which obviously cause significant


damage and loss of life.  Since there is often little time between the initial development of


the parent mesocyclone and the spawned tornado, it becomes critical to rapidly identify


these patterns and issue warnings. 


Using Doppler Weather RADAR, it is possible to detect these features in near


real-time.  However, in order to be able to detect these storms, we must have good


knowledge of exactly what we are looking for on the RADAR display. A key to this is


mastery over the theory of mesocyclone development.  This paper will present the latest


accepted theory on this subject.  Because this is such a potentially complicated topic, only


a brief overview of the theory will be provided.  It is essential to have as complete an


understanding as possible, in order to interpret the phenomena on the RADAR display


with optimum efficiency.


Our focus will be on the WSR-88D (Next Generation Weather RADAR – or


NEXRAD).  This is due to the fact that this S-band RADAR is well situated across the


continental US in a fairly dense network of mostly overlapping RADAR sites.  Further,


the NEXRAD system (3 GHz , 10 cm, PRF 318 to 1000 Hz) possesses good


characteristics to detect these phenomena.


Pattern recognition of mesocyclones - in both reflectivity and radial velocity


fields - will be discussed. Reflectivity and radial velocity are the two main parameters


that most meteorologists will consider first on the NEXRAD.  The use of reflectivity is


well documented, and as such only a brief period of time will be spent here, summarizing


main points.  Radial velocity data will be studied closer, since this data gives keen, quick


insight into the development of rotation in a thunderstorm.


After this is complete, our attention will shift to two different cases that


demonstrate the detection of mesocyclones.


Having looked at the pattern recognition aspects of mesocyclones in NEXRAD,


we will then pay particular attention to the use of computer based algorithms in the


NEXRAD system to aid in the automated detection of mesocyclones.  As a new version


of NEXRAD software is now coming out (called “Build 10” – a new NEXRAD


software/data processing series- from the NEXRAD Operational Support Facility, or


OSF, in Norman Oklahoma), it becomes even more important that ever to understand


what the NEXRAD is telling us.  I will demonstrate that there are significant changes in


the algorithms for tornado detection from the NEXRAD OSF in Build 10




            To start this discussion, let us first obtain an initial definition of  the mesoscale


cyclone, or “mesocyclone”.  Rogers and Yau [1976] defined it as a “…horizontal


circulation about 10 km across with values of vertical vorticity in the order  of 10-2 /sec”.


The mesocyclone is a mesoscale feature of the supercell thunderstorm.  Glass and


Przybylinski  [1994] provided the following background on mesocyclones:


From 1971 through 1977, the National Severe Storms Laboratory (NSSL) conducted an extensive study on the evolution of mesocyclones across Oklahoma [Burgess et all, 1982].  Over forty mesocyclones detected within 150 km during the middle and late spring seasons were analyzed to determine mesocyclone characteristics.  {This study helped determine some of the vortex recognition criteria used to establish persistence - we will discuss this later.}  Their study showed that the  majority of mesocyclones (76%) were comprised of one core during their lifetimes, while a minority of circulations went on to produce second and successive cores.  Both single and successive cores evolved through  a three stage lifecycle (i.e. Organizing, Mature, and Dissipating Stages) in which a vortex exhibits unique velocity characteristics during each stage.  Their results indicated that during the Organizing  Stage, mesocyclones expand upward and downward  from mid-level beginnings (4-6 km or 13,000-19,500 ft).   The Mature Stage begins when the mesocyclone base extends either to the surface or the lowest level above the ground.  At this point, the mesocyclone may exceed overall heights of 8 km (26,000 ft) and exhibits its strongest intensity.  The signatures which define the  Dissipating stage include a significant drop off in mesocyclone depth and weakening rotational velocities.  Of the total number of circulations surveyed, 95% of them produced some type of surface damage while nearly 60% of them were associated with tornadoes.” [Glass and Przybylinski, 1994]


Glass and Przybylinski  [1994] also indicated that follow on studies (conducted since the


earlier work from the 1970’s began) show that an average of 50% of all mesocyclones


tend to produce tornadoes, and that 90% produce some type of severe weather (tornadoes,


high winds, or hail).  Almost all tornadoes, on the other hand, are produced from parent


mesocyclones.  In one of the studies conducted by NSSL [Brown, Lemon, and Burgess,


1977],  indications were made that in one vigorous case, it took 25 minutes for a clear


mesocyclone - detectable by a Doppler RADAR-  to actually produce a tornado.  This has


clear implications for the operational forecaster that is attempting to provide warnings to


the general public: detect the parent mesocyclone, and you have a good chance of at least


getting severe weather, if not a tornado.


            This overview is a good first look at the mesocyclone, but we will need a


somewhat better physical description of the mesocyclone (and the supercell


thunderstorm, for that matter) if we hope to actually be able to recognize such a feature


on our NEXRAD PPI.  Therefore, a short overview of the supercell and its evolution will




            The first step in this process is the establishment of a thermodynamically unstable


environment with strong vertical wind shear.  Figure 1 below demonstrates the shear

Figure 1: Shear Environment for a Supercell


environment.  This figure starts our consideration.  As the shear environment develops,  


regions of horizontal vorticity are induced.  The line of overall horizontal vorticity is


shown by the line connecting the two plus signs in the figure.  As these lines of vorticity


are advected into regions of upward vertical motion, these horizontal lines of vorticity are


deformed.  See figure 2 for a visual demonstration of this effect.  In figure 2, the lines of


horizontal vorticity come under the influence of a ‘bump’ – this is the region of


instability.  This unstable air rises, causing the entire vortical line to be raised and


deformed to match the rising air.


This deformation causes the positive horizontal vorticity lines to tilt into


two vertical lines of vorticity – one line is positive, one is negative.  As the unstable air


continues to ascend, and the cumulonimbus cell develops, the pair of vertically oriented

Figure 2:  Deformation (tilting) of lines with horizontal vorticity into vertical vorticity. [figure from Rotunno, 1993]


vorticity lines form two rotating cells.  One is cyclonic, and the other is anti-cyclonic. See


figure 3 to see this graphically.  These cells will both contribute to the vertical instability

Figure 3:  Formation of mesoscale cyclonically and anticyclonically rotating cells. [figure from Rotunno, 1993]


of the overall cloud system.  As the cloud continues to grow, the moisture in the cloud


will inevitably condense and fall out as rain.  As the rain continues, it will drag down air


with it, causing another deformation in the line identified by the (a) in figure 3.  This will


lead to the birth of two new cells of positive and negative vertical vorticity, as shown in


figure 4.  As the thunderstorm system continues to move (in the overall direction of the


upper level wind field), the storm will continue to draw in moisture (which will help


continue the rain), and it will encounter (or run into) additional lines of horizontal


vorticity, which will keep the rotational energy going in the vertical vorticity cells.


Eventually, these two factors will cause the storm to split in the middle, giving rise to two


new thunderstorms, similar to the ones seen in figure 3 previously.  This could, in theory,


could simply continue on, with new cells forming and splitting.  However, we will see


that this would only be possible if the vertical shear pattern remains as shown in figure 1.

Figure 4:  Formation of the new set of mesoscale cyclonic and anticyclonic cells due to rainfall activity. [figure from Rotunno, 1993]


Note also the following feature associated with the vertical profile:  as the wind blows on


the windward side of the thunderstorm, a relative maximum of pressure is induced,


whereas a minimum of pressure is induced on the leeward side of the cell.  This is


portrayed in figure 5 below.  The nature of the mean shear is repeated and emphasized by


the hodograph in the bottom right of this figure. Note that this situation is, in fact, no


different from figure 3 at all – it is the same situation.  This serves to continue the process


that we have already discussed. 

Figure 5:  Formation of the microscale relative high pressure on the windward side of the storm, and the microscale low pressure on the leeward side. [figure from Rotunno, 1993]


            This above situation will change dramatically when we introduce a new vertical


wind distribution.  This is depicted below in figure 6.  Note the new hodograph – we now


have a very different pressure gradient situation acting on the cloud at different levels. 


Note that a new feedback mechanism takes place.  Take a closer look at figure 6: there is

Figure 6:  Formation of the microscale relative high and low pressure regions with a cyclonically curved hodograph. [figure from Rotunno, 1993]


a new feature here:  on the left side of the storm (this is on your right as you look at the


figure) a relative microscale high pressure is built up in the upper portion of the storm,


whereas a relative low pressure region is developed below.  This will discourage upward


vertical motion on the left side of the storm.  In the previous case, this was the side with


the anticyclonically rotating cell.  On the right side, the exact opposite takes place: a


relative microscale high pressure region lies at the lower level, whereas relative


lower pressure is above.  This encourages additional upward vertical motion.  This is


taking place on the side with the cyclonically rotating vorticity cell.  As this process


continues, the cyclonically rotating cell is strengthened, whereas the anticyclonically


rotating cell is suppressed.   Note that this is the case for a cyclonically curved


hodograph.  If we were to have the reverse case (an anticyclonically rotating hodograph),


then we would see the anticyclonically rotating cell become dominant.


            The net effect of all this is shown below in figure 7, which depicts our


thunderstorm system.  This is otherwise known as the classic Supercell.  The cyclonically


rotating vertically oriented vortical circulation is the classic mid-level mesocyclone.  It


normally forms in the middle levels of the storm, somewhere between 4-6 km on average.


It is reasoned by many that the mesocyclone will continue to build both upward and


downward through the supercell. As it builds downward, it will, at least 30-50% of the


time, develop into a tornado.  The development of the mesocyclone is further aided by 

Figure 7:  The classic tornadic supercell. [figure from Rotunno, 1993]


unstable air rising in the cell.  As the air rises vigorously, the rotational column is


stretched in the vertical.  This will serve to decrease the moment of inertia of the column


of rotating air.  Of course, from physics, we will remember the conservation of angular


momentum, which requires that L=I . w  remain constant throughout (L is the angular momentum, I is the moment of inertia, and w is the angular velocity).  In short, given


the L must remain constant (let us ignore viscous effects for the short term), and that the


moment of inertia is decreasing (due to the vertical stretching process induced by the


upward vertical motion), we must then see an increase in the angular velocity.   This is


what leads to increasing mesocyclone intensity, and ultimate formation of the tornado. 


The upward stretched mesocyclone can reach far above and below the 4-6 km


level of its initial formation.  If it is, in fact, tornadic, then you will see the mesocyclone


reach from nearly the surface to 10 km or more.  Initially, the mesocyclone can be very


wide in diameter.  Typically, these diameters vary from 1-10 km, depending on the size


and strength of the cyclone [Stumpf et al, 1997].


            This concludes our discussion of the supercell and mesocyclone.  We now shift


the focus onto the aspects of RADAR detection of the mesocyclone.




            We will now look at the mesocyclone of the RADAR’s perspective.  What will


the RADAR see?  Let’s look at the supercell from the top down, as in figure 8.  The


image has a north arrow pointing upward on the paper, and the storm is moving to the


east-northeast.  The yellow region generally corresponds to a region of precipitation.  In


the forward part of the region of rain, we see the forward flank downdraft , or FFD (a

Figure 8:  A plan view of the tornadic supercell. [figure from Finley, 1990]



feature we would have expected if we looked again at figure 1).  To the rear is the rear


flank downdraft, or RFD – a feature that is consistent with the microscale high pressure


we addressed earlier.  The arrows are the near surface streamlines that will clue us into


the expected nature of our isodops (lines of equal Doppler velocity on the RADAR


display) at low RADAR antenna elevations.  A further inspection of the precipitation area


(see figure 9) gives us some good hope as to the expected reflectivities we will expect to


observe in the supercell.

Figure 9:  A plan view of hydrometer type and precipitation distribution in the tornadic supercell. [figure from Finley, 1990]




Let’s take another look at the yellow region in figure 8.  Here, we have a small,


cyclonically curved region in the southwestern part of the yellow precipitation band.  This


is called the HOOK ECHO, and it locates our mesocyclone very nicely.  A real world


example is provided below in figure 10.


Figure 10:  A PPI view of an actual tornadic supercell with classic hook pattern. The red triangle denotes a tornado as detected by Doppler RADAR - This storm did have a confirmed tornado [information and figure from the National Severe Storms Laboratory home page, 1997]



Note the reflectivities start high in red (55 dbZ +) and decrease through orange, yellow,


green, blue, and finally black for the lowest return.  This pattern of reflectivity is


consistent with the pattern of precipitation and hydrometeors as seen in figure 9.





The problem with the hook echo is that tornadoes – especially F0 / F1 strength –


are not always going to exhibit the classic hook echo pattern.  The radial Doppler velocity


product can, on the other hand,  tell us a great deal about the internal working of the


supercell, even giving us a quicker detection time for mesocyclones.  In general, with the


mid-level mesocyclone, we would expect the following radial wind distribution, as seen


in figure 11.  This is the RANKINE COMBINED VORTEX, and it represents the

Figure 11:  The Rankine Combined Vortex.  [figure from Lemon et al, 1978]

theoretical distribution of winds in the horizontal plane, associated with the mesocyclone
(note that this is a similar wind model for the hurricane).  The Rankine Combined Vortex
(or RCV) is made up of two regimes  - one is the solidly rotating core, where the winds
are given by the equations below:



       V  =C1                                 C1=Vmax/R



where R is the radius of the vortex core, r is less than R, V is tangential velocity, and


Vmax is the maximum tangential velocity  (C1 is obviously meant to be a constant,


which means that the solidly rotating core increases linearly in absolute tangential


velocity with distance from the center of the core.).  Outside the solidly rotating core,

there is another flow region, which is referred to as the potential vortex flow.  This is


given by the equations below:


       Vr  =C2                               C2=VmaxR


The variables are the same as above, with C2 also being a constant  [These equations


come from the work of Lemon at al, 1978].  Lemon  et al [1978] also listed four initial


criteria for  identifying a signature as a true mesocyclone: 



“1) There is significant tangential shear of radial velocity in a quasi-horizontal plane where the sum of the angular diameter and elevation angle of observation is less than 30 degrees

2) Tangential shear persists for at least half the period required for one vortex revolution.

3) Vertical extent of the shear pattern exceed horizontal diameter

4) Qualitative shear pattern is invariant during a viewing angle change approaching or exceeding 45 degrees.”

[Lemon et al, 1978]



This represents an initial set of identifying conditions needed to identify a velocity pattern


as a mesocyclone.  These actually do not represent a list of sufficient conditions, but they


are necessary conditions.


            An example from a real case is given in Figure 12.   Here, the reflectivity data is


the left, and velocity data is on the right.  Look at the velocity data near Arcadia


Oklahoma: there is significant wind shear taking place in both the top panel (elevation


1.8 degrees) and even more so on the velocity panel on the lower right (0.4 degrees),


again near Arcadia, to the southwest (or to the northeast of Oklahoma City).  This is the


clear cut example of the tornadic mesocyclone.



Figure 12:  A four panel PPI view of an actual tornadic supercell with classic mesocyclone velocity pattern. The velocity data is on the right.  Note the contrast of inbound winds (green) and outbound winds (red) – a sure sign of a mesocyclone.  The red triangle (left two panels) denotes a tornado as detected by NEXRAD Doppler RADAR algorithms  - This storm did have a confirmed tornado on 13 Jun 1998 [information and figure from the National Severe Storms Laboratory home page, 1997]

Figure 13:  A four panel PPI view of another actual tornadic supercell with weaker mesocyclone velocity pattern.  The RADAR site is due west of this location. The velocity data is on the right.  Note the contrast of inbound winds (green) and outbound winds (red) due west of the Isle of Wight (near Norfolk Va)– a sure sign of a mesocyclone.  Note that the  NEXRAD Doppler RADAR algorithms didn’t identify a tornado here, although this storm did have a confirmed F1 tornado on 12 Jun 1996 [information and figure from the National Severe Storms Laboratory home page, 1997]

            Another example from a real case is given in Figure 13.   Here, the reflectivity


data doesn’t seem to give a clear hook pattern, yet there is a contrast of velocity near the


Isle of Wight (west of Norfolk Virginia).  This velocity contrast is, in fact, another


mesocyclone, like figure 12, and an F1 tornado was associated with this event.  This


example shows that reflectivity alone is insufficient to detect tornadic cells. 


            We have paid much attention to the mesocyclone in the horizontal, but let us take


a moment to remember that the mesocyclone is also a three dimensional feature.  The


circulation may appear impressive initially at the middle levels, but unless the feature


builds vertically (up and down) then the feature has not attained full mesocyclone status. 


The only way to do this with the RADAR is to scan the feature at different elevations,


and observe the circulation at more than one level.  Both figures 12 and 13 have this –


both features show two different elevation scans, and the mesocyclone is evident in both


scans in each case.  Doing this will increase our confidence that a mesocyclone is


forming (or already exists).  Keep in mind that many transitory features (shears and


circulations) have been observed in shallow layers – often in the boundary layer.  Again,


as mentioned earlier, we will need to look for significant vertical depth to this signature. 


What is ‘significant vertical depth’?


            The proposed answer was given by Keighton and Medlin [1995] as follows: 


“….[the] depth of the rotational signature must be at least comparable to the  diameter of the mesocyclone core.  In other words,  a broad rotational core signature (say 5 nm) would need to be roughly 5 nm deep (approx. 30,000 ft) for confident identification as a bona fide mesocyclone.  The depth criteria were further modified for operational use ... in the late 1970’s …to allow for a minimum of about 50% of the core diameter, as long as it was not less than 3 km (10,000 ft) in depth. Keep in mind that this particular depth criterion was established from a data set consisting primarily of Plains supercells.” [Keighton and Medlin 1995]


Another note is that the mesocyclone should have some time continuity.  The


rule of thumb suggested by Keighton and Medlin [1995] was that the mesocyclone


should be evident for two volume scans for initial mesocyclone detection.  However,


Keighton and Medlin [1995] also pointed out that …..


“…an exception to the time continuity criterion occurs in the event that the signature is already strong and deep when discovered.  In this situation, immediate action is justified! Remember,  those mesocyclone cores that are characterized by the highest rotational velocities and exhibit a tendency to build downward with time pose the highest tornadic threat!” [Keighton and Medlin 1995]


A final note concerns a limitation of the RADAR.  Since it samples at discrete


time intervals and discrete volume sizes, there may be situations where the signature


changes rapidly from volume scan to volume scan.  This can be a very significant factor


at longer ranges.  In reality, the feature may be steady state, but this can become


challenging to the analyst without experience.  Caution is the key word here!



            The WSR-88D, otherwise known as the Next Generation Weather RADAR


(NEXRAD) is a valuable tool for meteorologists, since the RADAR Product Generator


(or RPG) within the NEXRAD system can digitally run numerical algorithms needed to


objectively estimate severe thunderstorm activity.  One such application is called the


Mesocyclone Detection Algorithm (or MDA), which is currently used in the RPG’s


existing software.  The current software version is called ‘Build 9’; at the time this paper


is being written, the National Weather Service is preparing to implement a new software


version called ‘Build 10’.  It is significant to note that no change in the MDA is expected


for Build 10, although the method of detecting tornadoes is changing quite a bit. We will


focus on the MDA here, and briefly mention both the Tornado Vortex Signature (TVS in


Build 9) and the Tornado Detection Algorithm (TDA) in Build 10.  Please note that


almost all this section of this paper, section 4, is obtained from the document from the


NEXRAD Operational Support Facility (OSF) in Norman Oklahoma, entitled “WSR-88D


Build 10 Training” including all figures.


            In Builds 9 and 10, the MDA begins with a careful inspection of Doppler radial


velocities in a given RADAR antenna angle, and looks at adjacent range bins, or range


gates.  The RPG, a specialized digital computer system, determines groups of lines at a


constant range from the RADAR that exhibit increasing Doppler velocity. This means


that the processor looks for a group of velocities that go from negative values of velocity


to positive values of velocity with increasing azimuth (the RPG doesn’t look for the


opposite case as a mesocyclone, but the opposite case will be identified as a significant


region of shear).  What this all means is that the NEXRAD only looks for cyclonically


rotating mesocyclones.  Once the RPG identifies a region as having a positive gradient of


radial velocity from negative to positive, it calls such a grouping a ‘run’ of increasing


velocities with increasing azimuth.  This is then called a ‘pattern vector’.  The MDA will


attempt to identify all the pattern vectors it can manage, as shown below in figure 14.


Figure 14:  A plan view of the pattern vectors defined by the MDA in the  NEXRAD RPG are highlighted in blue.  Note the location of the RADAR Antenna (the RDA) below [figure and information from the NEXRAD OSF ‘WSR-88D Build 10 Training’ Document ]


Once the pattern vectors are identified, calculations are made to determine if certain


minimum shear and momentum criteria are met.  If these criteria are not met, then the


pattern vector in question is discarded. See figure 15 for an example of this.

Figure 15:  Rejected pattern vectors are highlighted in green. [figure and information from the NEXRAD OSF ‘WSR-88D Build 10 Training Document ]

Once this takes place, another algorithm is run. Called the ‘Threshold Pattern Vector


(TPV)’ in NEXRAD, it is user definable.  This algorithm ensures that a minimum number


of correlated pattern vectors are present (normally this is set to 10) to have the system


lock in a feature as a two – dimensionally correlated feature.  See figure 16 below for an




Figure 16:  Correlated pattern vectors are identified as correlated 2-D features, highlighted by the thick black line. [figure and information from the NEXRAD OSF ‘WSR-88D Build 10 Training’ Document ]


Once this process in complete, this feature must be correlated with other features in the


vertical.  NEXRAD scans in other elevations, for example, and looks to see if there are


other 2-D features that correspond to this feature you see above in figure 16.  See figure


Figure 17:  Vertically correlated

2-D features are identified and correlated as 3-D features, highlighted by the grey areas. [figure and information from the NEXRAD OSF ‘WSR-88D Build 10 Training’ Document ]



17 to get a feel for what the NEXRAD is scanning.




It is worthwhile to note that NEXRAD does allow for one missing 2-D feature in one of


the elevation angles to allow for the possibility of range/velocity aliasing.


Once the 3-D correlated feature is defined, it is labeled a ‘mesocyclone


circulation’ with a circle on the NEXRAD PUP (Principal User Processor) display, as


long as minimum horizontal to vertical aspect ratio is maintained (see the previous


section for a discussion on this).


            For the Tornado Vortex Signature (TVS in NEXRAD), this basically takes


advantage of all the work done by the MDA. In order to have a TVS, there must be two


elevation slices that meet a certain threshold  shear value.  This shear value is determined


within a given slice by taking the maximum inbound and maximum outbound radial


winds, as shown in figure 18.


Figure 18:  Maximum  inbound and outbound winds are highlighted in green and red for the TVS calculation. [figure and information from the NEXRAD OSF ‘WSR-88D Build 10 Training’ Document ]



            The actual value of this shear is definable by the user in the TVS Threshold Shear


(TTS) algorithm in NEXRAD.  If the minimum shear is met, a TVS is located on the


PUP display with either a red triangle (tornado) or a yellow triangle (ETVS – elevated


TVS, or funnel cloud). The key thing to remember is that the MDA MUST DETECT A




mesocyclone identified by the MDA, the TVS (Build 9) will not execute.


            The new algorithm for Build 10, the Tornado Detection Algorithm (TDA) works


differently. First, it is a totally independent algorithm – it does not need the MDA to


identify a mesocyclone.  There are three steps to the TDA.


            In the first step, the TDA identifies pattern vectors in a manner similar to the


MDA.  The difference is that the shear and momentum thresholds are calculated as the


pattern vector is defined. This is seen in Figure 19. One other item to note here is that the


TDA has a high shear/momentum threshold.  Only CYCLONIC shears are identified.


Figure 19:  Maximum  inbound and outbound winds are highlighted in pink for the TDA calculation. The additional pattern vectors (in blue) represent the MDA features, for comparison.  [figure and information from the NEXRAD OSF ‘WSR-88D Build 10 Training’ Document ]



            In the second step, the pattern vectors are correlated into 2-D features.  Two


things are done here: One, there must be a minimum of three (3) pattern vectors  in


order to define a 2-D feature.  Second, TDA runs a symmetry test to ensure the feature


maintains a certain range of length to width ratio.  If this is accomplished, then a 2-D


feature is identified.  See figure 20 for an example of this 2-D depiction.


            Finally, the two dimensional features are vertically correlated, like the MDA

Figure 20:  The 2-D feature in the TDA is highlighted by the heavy black line. The additional pattern vectors (in blue, red, and green) represent the MDA/TVS features, for comparison.  [figure and information from the NEXRAD OSF ‘WSR-88D Build 10 Training’ Document ]


algorithm.  There needs to be a minimum of three (3) vertically correlated 2-D features in


order for the TDA to identify the overall feature as a 3-D feature (or, as either a tornado


or ETVS). Also, there may be a gap of one elevation where a 2-D feature is missing, in


order to account for range/velocity considerations. 


            Comparisons are underway for the TVS vs. the TDA.  Recent commentary (from


the October 1998 National Weather Association Meeting in Oklahoma City, which the


author attended ) seems to indicate that the new TDA algorithm has, in many cases, been


identifying phenomena in severe thunderstorms as tornadoes, when no tornadic activity


was actually observed. If this pattern holds true, then one would expect a higher false


alarm rate for tornado warnings in the near future.  One well identified advantage of the


TDA, though, was noted.  Apparently, when tornadoes DO occur, the TDA tends to alert


forecasters to the possibility of tornadoes faster that the older TVS did.  Therefore, there


are new strengths and weaknesses that must be further studied with regard to these


algorithms.  In the interim, forecasters should be cautious!



The WSR-88D has a tremendous capability to detect the many different features


of the supercell thunderstorm, using both reflectivity and radial velocity information


displayed on the Principal User Processor (PUP).  With this device, warnings can be


issued to the public in the event of severe/tornadic conditions.  The WSR-88D can also


provide automated detection of severe weather conditions using the Mesocyclone


Algorithm and the Tornado Detection Algorithm (TDA).  The new TDA has shown an


improved capability to detect tornadoes in shorter time than required by the older


Tornado Vortex Signature (TVS) from previous NEXRAD software builds, although


many in the operational community have already commented that the new TDA will tend


to identify regions as tornadic when, in fact, they are not exhibiting tornadic features. 


This will certainly lead to a higher false alarm rate for operational weather forecasting


organizations, such as the National Weather Service, in the near future. Care must be


exercised by operational forecasters to carefully analyze RADAR data by manual


inspection to help prevent false alarms, and to also pick up on tornadic cases that the


TDA might miss in the future.



Burgess, D.W., 1998: WSR-88D Build 10 Training, WSR-88D Operational Support Facility, Norman OK


Finley, C. 1990:  Technical Training: Doppler Weather Radar: Thunderstorm Morphology and Dynamics, 27 August 1990,  Chanute Technical Training Center (ATC), 3350 Technical Training Group, Chanute Air Force Base, Illinois.


Glass, F. H., and R. W. Przybylinski, 1994: Operational Considerations for the Detection of Mesocyclones at Long Ranges. Postprints, The 1st WSR-88D User’s Conference, Norman, OK, WSR-88D Operational Support Facility, Norman OK ,pp. 427-438.


Keighton, S,and J. Medlin, 1995:  The WSR-88D Operator’s Guide to Mesocyclone Recognition and Diagnosis, WSR-88D Operational Support Facility, Norman OK


Lemon, L. R.,  D. W. Burgess, and R. A Brown, 1978: Tornadic Storm Airflow and Morphology Derived from Single-Doppler Radar Measurements. Mon. Wea. Rev., 106, 48-61.


Rogers, R.R, and M.K. Yau, 1976: A Short Course in Cloud Physics, Pergamon Press Canada, Ltd., Suite No. 271, 253 College Street, Toronto, Ontario, Canada, M5T 1R5, 3rd edition, Copyright © 1989.


Rotunno, R., 1993: Supercell Thunderstorm Modelling and Theory. Geophysics Monograph 79, The Tornado: Its Structure, Dynamics, Prediction, and Hazards , 57-73.


Stumpf, G. J., A. Witt, E. D. Mitchell, P. L. Spencer, J. T. Johnson, M. D. Eilts, K. W. Thomas, and D.W Burgess, 1997: National Severe Storms Laboratory Mesocyclone Detection Algorithm for the WSR-88D, National Severe Storms Laboratory, Norman, Oklahoma, Weather and Forecasting, Vol. 13, pp. 304-326.


Author Unknown, 1998: WSR-88D Build 10 Training, National Weather Service, NEXRAD Operational Support Facility, Norman Oklahoma, available at





Andra, D., Preston, V., Quetone, E., Sharp, D., and Spoden, P., 1994: An Operational Guide to Configuring a WSR-88D Principal User Processor, Operations Training Branch, Operational Support Facility.


Burgess, D. W., and L. R. Lemon, 1991: Characteristics of mesocyclones detected during a NEXRAD test. Preprints, 25th Int'l Conference on Radar Meteorology., Norman, Oklahoma, American Meteorological Society, 39-42.


______, R. L. Lee, S. S. Parker, S. J. Keighton, and D. L. Floyd, 1995: A Study of mini supercells observed by WSR-88D radars. Preprints, 27th Conference on Radar Meteorology, Vail, Colorado, American Meteorological Society.


Lee, R. L., 1996: Improvement of the WSR-88D Mesocyclone Algorithm, WSR-88D Operational Support Facility, Norman, Oklahoma.


Stumpf, G. J., and C. Marzban, 1995: NSSL Build 2.0 Mesocyclone Detection Algorithm ( MDA2), National Severe Storms Laboratory, Norman, Oklahoma.


Wicker, L. J., and L. Cantrell, 1996: The Role of Vertical Buoyancy Distributions in Miniature Supercells. Preprints, 18th Conference on Severe Local Storms, San Francisco, CA, American Meteorological Society, 225-229.


Brooks, H. E., C. A. Doswell III, and R. B. Wilhelmson, 1994: The role of midtropospheric winds in the evolution and maintenance of low-level mesocyclones. Mon. Wea. Rev., 122, 126-136.


_____, _____, M. T. Carr, and J. E. Ruthford, 1996: Preliminary analysis of soundings from VORTEX-95. Preprints, 18th Conf. on Severe Local Storms, San Francisco, CA, Amer. Meteor. Soc., 133-136.


Burgess, D. W., R. L. Lee, S. S. Parker, S. J. Keighton, and D. L. Floyd, 1995: A Study of mini supercells observed by WSR-88D radars. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Soc., 4-6.


Grant, B. N., and R. Prentice, 1996: Mesocyclone characteristics of mini supercell thunderstorms. Preprints, 15th Conf. On Wea. Anal. and Forecasting, Norfolk, VA, Amer. Meteor. Soc. , 362-365.


Lee, R. R., 1995: Improvement of the WSR-88D Mesocyclone Algorithm. Documentation available from the WSR-88D Operational Support Facility/Application Branch, Norman, OK, 73069, 44 pp.


Marzban, C., and H. Paik, 1996: Various Statistical Analyses of MDA Data. Paper currently in press. (Draft available from the NSSL, Norman, OK, 73069)


Mitchell, E. D., 1995: An enhanced NSSL tornado detection algorithm. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Sci., 406-408.


Rasmussen, E. N., J. M. Straka, R. Davies-Jones, C. A. Doswell III, F. H. Carr, M. D. Eilts, and D. R. MacGorman, 1994: Verification of the origins of rotation in tornadoes experiment: VORTEX. Bull. Amer. Meteor. Soc., 75, 995-1006.


Stumpf, G. J., and C. Marzban, 1995: NSSL Build 2.0 Mesocyclone Detection Algorithm (MDA2), Final Documentation. Available from the NSSL, Norman, OK 73069, 76 pp.


_____, _____, and E. N. Rasmussen, 1995: The new NSSL Mesocyclone Detection Algorithm: A paradigm shift in the understanding of storm-scale circulation detection. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Sci., 208-210.


Trapp, R. J., and E. D. Mitchell, 1995: Characteristics of tornadic vortex signatures detected by WSR-88D radars. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Sci., 211-212.