Verification of the MDA Using DOW Velocity Data

Radar Project

METR 5613-001

Thomas Jones

April 26, 2002

1. Introduction

This research attempts to utilize the Doppler-On-Wheels (DOW) mobile radars in a verification experiment against algorithm output from the WSR-88D radars (Wurman et al. 1996). The WSR-88D algorithm to be studied is the NSSL mesocyclone detection algorithm (MDA) (Stumpf et al. 1998). The MDA uses Doppler velocity data acquired by the WSR-88D radar and attempts to derive the location and attributes of storm-scale vortices from these data. Unfortunately, there has been no attempt to verify whether the vortex detections produced by the MDA are indeed real and accurate. An independent source of velocity data is required to accomplish this. The DOWs can provide this independent data source and make possible for the first time a study in the true ability of the MDA to detect mesocyclones and their attributes.

The MDA produces several mesocyclone detection attributes that can be analyzed. These include gate to gate velocity difference, rotational velocity difference, shear, and mesocyclone depth. However, the accuracy of these attributes is somewhat suspect. Several factors can cause false and misleading mesocyclone detections. First, the WSR-88D velocity dealiasing algorithm often incorrectly dealiases small regions of data causing the reported velocity vectors to be opposite in direction than to what they really are. This in turn can cause the MDA to detect a mesocyclone where there is not a true vortex. Even given accurate velocity data, the MDA often fails to account for storm motion and vertical association characteristics of a mesocyclone. This can lead to incorrect values for a mesocyclonesís attributes even if the detection location is correct. Using the DOW to collect comparable data may provide greater insight into the overall accuracy of MDA detection attributes and enhance the ability to compare those attributes to additional phenomena.

The MDA in its current, operational form only detects cyclonic vortices. The rational for this is that cyclones are much more common in the Northern Hemisphere than anticyclones, and that most anticyclonic detections would be artifacts of true cyclonic detections (Fig. 1). However, the degree to which this assumption is correct has never been tested using WSR-88D data. The MDA can be "reversed" so as to operate in anticyclonic mode and recent research suggests that the number of anticyclonic detections is approximately equal to the number of cyclonic detections for the same event. It is difficult to believe that all these anticyclonic detections are indeed artifacts. Some must represent true anticyclonic circulations and it is hoped that this work can provide a rough estimate of the percentage of anticyclonic detections that are indeed real not just artifacts. (Detections which were caused by faulty dealiasing or other radar problems were discounted prior to proceeding with that analysis). Then, if a significant percentage do prove real, further analysis of their importance in severe weather could then be studied.
 
 

Figure 1. Schematic of the "artifact" mesoanticyclone detection showing the true mesocyclone detection, C, with its associated toward and away velocity vectors. Away from the center of the mesocyclone the toward and away wind vectors are naturally less as velocity decreases as the distance from the center of the mesocyclone detection increases. Due to this anticyclonic shear, A, is created on the sides of the primary cyclonic circulation. The MDA sometimes interprets this anticyclonic shear as a mesoanticyclone when in fact it is not one.
 
  By comparing the velocity characteristics of both the WSR-88D and the DOW data with algorithm output, this research hopes to determine the skill of the MDA using an independent data source. While the amount of DOW data used in this initial comparison is very limited and not nearly enough to produce conclusive results, it is hoped that it can provide some idea as to MDA skill. This could then be used as the basis for future research.

2. Data Acquisition

In order to acquire DOW data that is comparable to WSR-88D data, DOW data was to be collected in modes that used similar range bins and unambiguous velocities as the WSR-88D. However, the DOW settings were set so that higher gate resolutions could be received to ensure that any features seen by the WSR-88D are also seen by the DOW. Due to the setting used by the DOW, the DOW had significantly greater resolution than the WSR-88D radar enabling the DOW to be positioned 10-20km away from the target area and still get the resolution required. Getting too near a rotating event with the DOW would not be useful for this research as the DOW would mostly sample the atmosphere below the lowest elevation scan of a WSR-88D. To account for the fact that the WSR-88D has a five to six minute interval between volume scans, the DOW sampled the storm events at a higher frequency (e.g. two minutes per volume scan) so that a near-time match between DOW data and WSR-88D data could be found. (Also, the clock on WSR-88D radars is not set to any specific standard and thus may be several minutes off the "true" time as well.) In all, approximately one hour of DOW data per event was analyzed in this initial survey.

In order to successfully accomplish the goals of this research, some rotating convection is required. Ideally, a rotating supercell thunderstorm would be used in this algorithm verification research. A supercell with a strong-long lasting mesocyclone would result in strong mesocyclone detections from the MDA and would be easily observable by a DOW. The location and attributes of the DOW velocity data could then be easily compared to the MDA output. An isolated supercell simplifies this comparison, but is not required for the success of this project. In fact, this research does not even require strong storm vortices to be successful. Any vortex that is observable by the WSR-88D and within ~200 km of a WSR-88D radar could then be observed by a DOW positioned at a much closer range. Ideally, only mesocyclone detections less than 100 km away from a particular WSR-88D should be used as the horizontal resolution of the WSR-88D beyond that range becomes quite poor. However, given the time constraints inherent in the project, this ideal situation was not used as a requirement. To be detected by the MDA a vortex must have a gate to gate velocity difference greater that 10 ms-1. Vortices this weak are common in many types convection such as squall lines and gust fronts. Since the threshold for detection by the MDA is so low, it is expected that any true circulations detected by the MDA should be observable with the DOW especially given the DOW dataís finer temporal and spatial resolution.

Using these techniques, the DOW was used to collect data for one event. This event occurred on 9 March 2002 between 3:00 and 5:15 Zulu time. The corresponding WSR-88D data used to run the MDA originated from the KTLX: Twin Lakes, OK radar1. Given the relative positions of the DOW and KTLX, the location requirements stated above were met by a significant margin. Further analysis of the event will be presented in the Results portion of this work. To supplement the data set collected above, an additional supercell case collected by DOW2 and DOW3 was also used during this analysis. This case occurred on 29-30 May 2001 and data between 22:00Z and 0:00Z was acquired. Since this case occurred with in the range of both the KAMA: Amarillo, TX and the KLBB: Lubbock, TX WSR-88D radars, data and algorithm output from both radars was initially analyzed. After a brief survey of the radar data available for this case, it was decided to use the KAMA data for algorithm output and DOW2 as the secondary source of radar data. Table 1 gives the settings used for DOW2 during the data collection process for both cases.

Table 1. DOW2 parameter settings for both data collection periods. Included for comparison are similar parameter values for the WSR-88D. Note that the given gate size for the WSR-88D is for velocity data only; reflectivity is recorded in 1 km gates.
 
 
  3. Analysis Techniques

The analysis began by finding mesocyclone detections from the MDA that occurred during the time and within the range of the DOW. The location of each mesocyclone detection was determined using the known lat-lon position of those detections provided by the MDA. These positions are only given to a precision of one kilometer and the accuracy is somewhat worse due to radar geometry issues. Thus, small spatial errors in the MDA detections may exist and must be taken into account. For each volume scan of WSR-88D data, the MDA defines the begin time of that volume scan as the time of occurrence for each mesocyclone detected during that volume scan. Unfortunately, this combined with the lack of a standard clock for the WSR-88D can cause temporal errors of several minutes from when the detection actually occurred. After taking into account these issues, the MDA detections for each event were ingested into Arc-View where that could be displayed in a spatially representative way. Please note the detections displayed are the 3-D detections which are made up of several 2-D components. Once the initial correspondence between these detections and the DOW data has been made, further analysis was undertaken using the 2-D detection information.

To make the actual comparison between vortices detected by the DOW and those detected by the MDA algorithm, the location of the centroid of a DOW detected vortex at a certain time will first be determined by hand. This process used the known spatial position of the DOW and the observed range and azimuth information. This process was complicated by the fact that the DOW was not pointed in the due North direction during the data collection period of each event. Since the DOW data display software (Solo II) displays the data in a truck relative North position, the data display must be mentally rotated so that it corresponds to true North2. If a correlation between a certain mesocyclone detection and DOW velocity data could not be found, further analysis as to the characteristics of the WSR-88D velocity data were analyzed. Possible causes of this discrepancy may be due to faulty dealising of WSR-88D velocity data and/or ground clutter interference. Under certain types of atmospheric conditions, it is possible for the MDA to produce many false detections. Due to higher resolution of the DOW data, the DOW should be able to resolve any vorticy that the WSR-88D can see. Thus, the DOW data will be used as a baseline for determine whether the circulations detected by the MDA are indeed real. Assuming a correlation is made between a MDA detection and a DOW vortex, the determination of which elevation angles to analyze that best correspond with the WSR-88D data must be made. To accomplish this, a custom program was designed that can take the 2-D information from each mesocyclone detection and its estimated range from the DOW and determine the elevation scan from the DOW that best corresponds with the MDA 2-D information. This process is carried out for both cyclonic and anticyclonic detections. For the purpose of this research, a mesocyclone detection is independent of time. Thus, detections of the same circulation during different volume scans of WSR-88D data will results in multiple mesocyclone detections and they will be counted as such.

Once the spatial and temporal correlation between MDA mesocyclone detections and vortices observed by the DOW, the maximum velocity difference will be calculated from the DOW data and compared to similar values derived from the MDA. Given the finer resolution of the DOW data, these values should be greater than or equal to those observed by the WSR-88D. If they were not, a more in depth analysis of both the DOW and WSR-88D for that particular case was undertaken in hope of determining the answer. Again, these answers include faulty dealiasing of either the DOW or WSR-88D data, and the possibility that the raw data itself for that case is somehow defective. For each DOW observed case, these comparisons were made for several vortices observed by both radars over the time span of the recorded data set.

In addition to the raw attribute analysis, the significance of the MDA detected meso-anticyclones was analyzed. Given an anticyclonic detection, DOW data for that detectionís time and place was analyzed to see if the anticyclonic detection was a result of weak anticyclonic shear induced by a larger cyclonic circulation or a true anticyclonic circulation. Both results were anticipated and the ratio of artifacts to real anticyclones was to be derived. However, in several cases ambiguity may exist as to the nature of the circulation as shown by the DOW data. If the nature of the circulation could not be placed into one of the categories about without a high degree of certainty, the corresponding WSR-88D detection was not used in creating the ratio. From this analysis, the percentage of cyclonic vs. anticyclonic detections will be determined along with the percentage of those anticyclonic detections that prove to be independent anticyclonic circulations.

Hopefully, these comparisons will be able to provide some insights, though not conclusive by any means, into the true abilities of the NSSL algorithm to detect storm scale vortices and their attributes.
 
 

4. Results

a. 8 March, 2002 Case:

This case is characterized best as a weak to moderate squall line along a strong cold front. In addition, it was associated with a strong outflow boundary that propagated 10 to 20 km ahead of the convection. Since this was far from the ideal case discussed above, the results garnered from this case where not expected to provide the previously anticipated results.

Figure 2. Arc View plot showing relative locations of the DOW, KTLX, and the vortex detections for the 9 March event. "ï" represents anticyclonic detections and "x" represents cyclonic detections.
 
  Figure 2 shows the location of the DOW during the time of data collection with respect to both the cyclonic and anticyclonic detections. Since the first hour of DOW data collected used a very coarse resolution, only DOW data between 4:00Z and 5:05Z on 9 March 2002 was used in the analysis. At first glance, it appears that a significant number of detections were associated with this case. During the space and time period for which DOW data were available, 19 cyclonic and 13 anticyclonic vortex detections were made. The greater number of cyclonic detections made was expected given that they are favored to form under most conditions in the Northern Hemisphere. However, further analysis of both the WSR-88D and DOW velocity data reveal that most of these detections are probably not true mesocyclones. From the cyclonic detections, only three could be directly correlated with DOW velocity data. The corresponding number for anticyclonic detections was one. The main cause for this discrepancy appears to be an algorithm fault in detecting multiple circulations along the outflow boundary, both of which are not associated with the convection.

While the DOW data does show small-scale vortices on the leading edge of the outflow boundary, they are small in scale (<1 km) and are not detectable over multiple scans. Since the WSR-88D data and is its associated algorithm output have a much coarser temporal and spatial resolution, and detection made by the algorithm under these circumstances cannot be verified. The MDA is not designed to handle outflow boundary situations, which is probably a large reason for its low performance in this case. Also, during several volume scans, significant velocity dealising errors occurred with the WSR-88D data. Since the MDA does cannot differentiate good velocity data from bad, several of the mesocyclone detections for this case appear to be a result of bad velocity data. Since the number of verifiable mesocyclone detections was so small for this case certain parameters such as the percentage of true anticyclonic vs. cyclonic detections were not calculated. Also, it appears that the sole "good" anticyclonic detection for this case which occurred at 4:35Z (88D time) is an artifact of a cyclonic circulation that was also detected (Figs. 3 and 4). One should notice that the cyclonic detection is associated with a strong (relatively speaking) gradient of incoming velocity while the anticyclonic detection is associated with a much broader gradient implying that there is no independent anticyclonic circulation. The exact nature of the circulation observed remains unclear. It does not appear to be a mesocyclone in the traditional sense as it is partially associated with the outflow boundary and only detectable at a few elevations. Also, this feature is not apparent a few minutes later suggesting it may have been a very quickly evolving event. Further statistics concerning the relationship between anticyclonic and cyclonic detections could not be made due to the small number of detections.

Figure 3. (above) WSR-88D 0.5° velocity data from the KTLX radar at 4:35 Z with anticyclonic (right) and cyclonic (left) MDA detections overlaid. Note that during this volume scan, several anticyclonic and cyclonic detections were made most along the outflow boundary (white). Also note the dealiasing error present to the SW leading to the false detection of mesoanticyclone #488. The white dot represents the location of the DOW during this time.
 
 

Figure 4. (left) DOW 1.7° velocity data from 4:38Z which reveals some shear to the SW of the radar. This shear or circulation appears to correspond with detections 453 and 563 respectively. The black circle represents the cyclonic circulation identified by the MDA as 563 while the anticyclonic shear on the opposite was detected as 453. Note that the gradient in velocity is much less on the anticyclonic side leading to the conclusion that 453 is an artifact of a true cyclonic circulation.
 
 

However, an attempt was made to verify the velocity difference attributes of the handful of mesocyclone detections that appear to be real given the DOW data. See table 2 for the numerical results. From this analysis, it appears that the MDA did an acceptable job at determining the velocity characteristics of the mesocyclones and mesoanticylones that were at least partially associated with convection. Unfortunately, this number only represents a very small percentage of the total number of both cyclonic and anticyclonic detections made during this event. For the rest, it appears from the DOW data that they are associated with small-scale vortices along the edge of the gust front. Most of these circulations have velocity differences that only marginally exceed the 10 ms-1 threshold used by the MDA indicating that the WSR-88D and the MDA would need to be quite sensitive in order to detect them. At first glance, the MDA appears to be detecting them. However, when the velocity attributes of the MDA detected mesocyclones are analyzed, they reveal circulations which have a maximum velocity difference of upwards of 50 ms-1. The velocity data from the DOW shows no such strong velocity signatures (Fig. 5). It is assumed that the nature of the outflow boundary causes difficulties in WSR-88D velocity dealiasing and decreases the noise to signal ratio leading to improper mesocyclone detections.

Figure 5. DOW 1.9° velocity from 4:47Z. Highlighted in red is a region of cyclonic shear with a *V of no more than 20 ms-1. However, the MDA detected a cyclone (#649) with a *V of over 50 ms-1. (See Table 2). Given the DOW velocity data from this and adjacent scans, there is no evidence of such a strong circulation. Thus, it appears that for this case the MDA grossly over estimated the strength of what was a true area of cyclonic shear.
 
 

Table 2. Table of selected MDA detections and their reported velocity differences with the corresponding DOW velocities listed for comparison for the 9 March case. Note that only three of the MDA detections could be located using DOW data. Most others could not be suggesting that the MDA with help from bad velocity dealiasing was not correctly detecting mesocyclones during this event. Since this event was outflow dominated and relatively weak the number of detections should be much lower. b. May 29, 2001 case

Due to the drought of rotating convection in OK during the month of March, it was decided to pull a more ideal case from the DOW archives in order to achieve more of the goals of this research. Thus, the supercell case from May 29, 2001 was chosen. Several supercells occurred on this day sponing several significant tornadoes. The storm the DOWs were observing as located roughly in between Amarillo and Lubbock, TX. The DOWs observed this storm for the better part of a day. This storm had with it a single, strong mesocyclone that sustained itself for the duration of the stormís life. Since there is only one true mesocyclone in which to analyze, this case proved much simpler to decipher and produced much more significant results. To simplify this research, only the observations from DOW2 between 22:12Z and 23:14Z were used in association with the KAMA WSR-88D data (Fig. 6).
 
 
 
 

Figure 6. Arc View plot showing relative locations of the DOW and the vortices detected during the 29 May event. "ï" represents anticyclonic detections and "x" represents cyclonic detections. Note that KAMA is not displayed as it is off the scale of this map to the North by an average of 90 km from the detections.
 
  During this observation period, 19 cyclonic detections were made along with 16 anticyclonic detections. Again, these numbers are consistent with what one would expect to detect in the Northern Hemisphere. In contrast to the squall-line case, nearly all of the detections could be associated with raw velocity data from both the WSR-88D and the DOW. A couple of the detections are partially associated with range folded WSR-88D data, but adjacent velocity information reveals that these detections are probably real. Nearly all of the cyclonic detections are associated with the large mesocyclone present in the supercell. During a couple of volume scans, multiple mesocyclone detections were made. These may represent the formation of new circulations as the supercell cycles during its life span (Figs. 7 and 8). The majority of the anticyclonic detections occurred in two storm relative locations. First, several detections where made just to the North and West of the large mesocyclone. Using the DOW data as a guide, it appears that these detects are indeed associated with some sort of anticyclonic circulation on the backside of the mesocyclone (Fig. 8). Second, some anticyclonic detections seem to occur within the precipitation core of the storm. The cause of these is unclear, but they do appear real given both the WSR-88D and DOW velocity data. Almost all of the anticyclonic detections appeared to be associated with an independent anticyclonic vortex. The few that could not be could not be associated with an "artifact" feature and were classified as ambiguous.

Figure 7. WSR-88D 0.5° velocity image from 22:43Z from KAMA. The supercell and its associated mesocyclone are quite apparent in this data. During this volume scan the MDA detects to cyclonic circulations (#146 and #449). Both of these detections are considered strong. See Table 3 for details. During this time the mesocyclone may have been "cycling" and with the old one weakening (#146) and a new on forming to the SW (#449). Also note worthy in the WSR-88D data is the low beam-width resolution in the region of the supercell. At this range each beam is over 1 km wide. Compare with corresponding DOW data in Figure 8.
 
 
Figure 8. DOW 19.0° velocity images from 22:44Z (left) and 22:51Z (right). Note the much greater resolution present in the DOW data than that of the WSR-88D data. The left image corresponds with the WSR-88D image shown in Figure 7. Note the main cyclonic circulation is very apparent and highlighted in red. The other circulation seen by KAMA is probably the smaller weaker circulation highlighted in green. The different orientation of the tow circulations as compared to the WSR-88D can be explained by the rapidly evolving nature of the event and time difference between the 2 scans. Also apparent at 22:44Z is a broad anticyclonic circulation on the NW side of the storm denoted in blue. This circulation appears real and independent of the cyclonic circulation but was not detected by the MDA until 22:53. As the storm approached the location of the DOW by 22:51Z, the DOW scan elevation was no longer high enough to reach the strongest areas of rotation as evidenced by the weaker velocity gradient present. (Actual weakening of the mesocyclone does not appear to be occurring on the WSR-88D data). Still, the ÆV measured by the DOW at 22:51Z is much greater than that seen by the WSR-88D showing how important fine resolution is.
 
  The velocity attributes of several cyclonic and anticyclonic detections were compared with velocity data from the DOW to determine how accurate the detection attributes are. See table 3 for the numerical results. One thing that is obvious is that the maximum velocity difference as detected by the MDA is noticeably less than the velocity difference observed by the DOW. The reason for this discrepancy appears to reside in the higher resolution of the DOW data. Most of these detections were made at approximately 90 km from the KAMA radar. At this range the beam-width of the WSR-88D pulse was approximately 1.5 km while the DOW was detecting the mesocyclone at a range of approximately 10 km resulting a beam-width of less than 0.2 km. When analyzing the DOW data, it appears that the areas of greatest velocity difference are small enough in size to not be apparent in the WSR-88D data. At this range the WSR-88D probably sees a multitude of different velocity measurements within the same gate. Some of these represent the true maximum winds while others have a significantly lesser value. Since the velocity estimate is averaged over the entire gate, one would expect the velocity values to be lower that those observed by higher resolution radar. From the DOW data, it appears that the mesocyclone maintained a velocity of 50 ms-1 or greater through to 23:00Z. After 23:00Z, the mesocyclone was too close to the DOW to be able to correlate information seen by the DOW to WSR-88D data. During this time the highest elevation scan made by the DOW was 19°, which given a mesocyclone range of less than 5km, fails to see data much above 1km. At the lowest elevation scan from the 88D radar, the height of lowest return was nearly 1.5km. The result of this is the lower velocity difference detected as the mesocyclone moved closer to the radar. The height at which maximum velocity difference occurred was not being observed. For the times when the correspondence could be made, it appears that the MDA underestimated the strength of the mesocyclone by an average of ~25 ms-1. For the anticyclonic detections, the DOW velocity estimates and MDA derived attributes were more in line though the MDA still produced an underestimate of the velocity difference. One possible explanation for the better performance appears to be the slightly broader and weaker characteristics of the anticyclonic circulations. Combined with less turbulent air around these circulations, it appears that the WSR-88D was better able to observe the true velocity characteristics of the mesoanticyclones.

Table 3. Table of selected MDA detections and their reported velocity differnces with the corresponding DOW velocities listed for comparison for the 29 May case. Note that with only two exceptions (both detections at 22:58Z) that the WSR-88D seriously underestimated the velocity difference of both cyclonic and anticyclonic detections. These exceptions are probably due to the DOW being too close to the rotation at that time. Also note that one broad anticyclonic circulation see by the DOW at 22:45 was not detected by the MDA. The circulation is visible in the WSR-88D velocity data and why the MDA did not pick it up is unknown.
 
  5. Conclusions

While the DOW data sets and storm events were not ideal to what was proposed, significant and important results were still obtained from this experiment. Analyzing the 9 March 2002 case revealed that the MDA had some major problems in its ability to detect mesocyclones in the vicinity of an outflow boundary. The MDA was not designed to handle outflow situations and this fact shows itself when the MDA detects many anomalous circulations along the outflow boundary. Worse, many of these anomalous detections are classified as strong, which when compared to the DOW velocity data is obviously not true. While this was not the goal of this research, analyzing this case seemed to verify that MDA is not capable of effectively dealing with outflow boundary situations.

The 29 May 2001 case provided significantly different results. First as the MDA was designed around detecting a single mesocyclone associated with a supercell, the percentage of "real" detections was expected to be much higher. Fortunately, it was. For both cyclonic and anticyclonic detections, all MDA detections could be verified from DOW velocity data assuming the DOW was not too close to see the appropriate feature. However, the MDA did produce velocity attributes that were much smaller that the DOW observed values. Given the range of the KAMA WSR-88D from the storm (90 km) and the much closer range of the DOW (10km), this was not unexpected. Still, the degree to which the stronger velocity difference got averaged out of the WSR-88D data was somewhat unexpected. Unfortunately, there is no practical solution to correct for this under estimation since it is mainly a result of the 1° beam-width of the WSR-88D radar. A smaller beam-width would require a larger WSR-88D antenna which is just not practical or oversampling of the beam which is not possible with the current, operational WSR-88D hardware.

Another surprising result of this research is that nearly all of the anticyclonic detections seemed to be associated with some sort of independent anticyclonic circulation as seem by the DOW data. No cases of "artifact" induced mesocyclone detections could be shown though a few where questionable. While it would be a stretch to call the anticyclonic features seen by the DOW full-fledged mesoanticyclones, they do appear to be real and may warrant further study. Hopefully, this research was able to provide a small insight into the real-life performance of the MDA. Even given this small sample of data, it is obvious that the MDA has a long way to go before it can reliably detect (and not detect) important circulations.

6. References

Stumpf, G.J., A. Witt, E. D. Mitchell, P. L. Spencer, J. T. Johnson, M. D. Eilts, K. W. Thomas, and D. W. Burgess, 1998: The National Severe Storms Laboratory mesocyclone detection algorithm for the WSR-88D.Wea. Forecasting. 13, 304-326.

Witt, A., M. D. Eilts, G. J. Stumpf, E. D. Mitchell, J. T. Johnson, and K. W. Thomas, 1998: Evaluating the performance of WSR-88D severe storm detection algorithms. The National Severe Storms Laboratory Tornado Detection Algorithm.Wea. Forecasting. 13, 313-318.

Wurman, J., J. Straka, and E. Rasmussen, 1996: Preliminary radar observations of the structure of tornadoes. Preprints, 18th Conference on Severe Local Storms. San Francisco, CA, Amer. Meteor. Soc., 17-22.
 
 

Notes:

1. The MDA (and other algorithms) were run on a Solaris Workstation using version 3.10 of the NSSL SSAP package and version 2.1 of the HIRES-dealising package. WSR-88D data was displayed using version 10.1 of WATADS on Rossby.

2. Dow data was processed using Solo II by NCAR and all velocity images shown in this paper where dealiased.
 
 

OU Home | Disclaimer | Copyright | Equal Opportunity | OU Web Policy