Robin J. Turner
National Weather Service Office
Vertically Integrated Liquid (VIL) values have been used, with some success, as an indicator of large hail during the past several severe thunderstorm seasons at NWSO Goodland, Kansas. Serious limitations of the VIL product, such as underestimation of values for strongly tilted or fast-moving storms, the large, day-to-day changes in the "VIL of the day," or the underestimation of the potential for thunderstorm severity during cool season events, have limited the utility of the VIL product at NWSO Goodland.
A new technique, developed at NWSO Tulsa, Oklahoma, may produce a greater amount of utility for using VILs to detect thunderstorm severity. Presented at the Southern Region WSR-88D Conference, the technique outlined by Amburn and Wolf in their paper VIL Density as a Hail Indicator (1995), may alleviate some problems associated with the VIL product.
Amburn and Wolf define VIL Density as:
where VIL is in units of kilograms per square meter and Echo Tops are in meters. The resulting VIL Density is given in units of grams per cubic meter (Amburn and Wolf 1995). A nomogram provided within the cited paper shows the following correlations:
VIL Density = 4.0 g/m3: 90 percent probability of large hail.
Since the WSR-88D produces echo top data in standard English measurements, it is useful to divide VIL by Echo Top, in thousands of feet, to produce the correlations below:
VIL Density = 1.22 kg/m2/kft: 90 percent probability of large hail
For this study, the previously discussed VIL Density values for large hail were compared to the VIL values likely (greater than 50 percent probability) needed for large hail in Oklahoma as presented in the WSR-88D Operator's course (Operational Support Facility 1993). The database used for comparison was obtained from Archive IV from the Principal User's processor (PUP) at NWSO Goodland, for the months of February through May of 1994. This database was selected since 1) it provided a large number of verified severe thunderstorm events as compared to the early portions of the 1993 and 1995 severe thunderstorm seasons, and 2) several low-topped severe thunderstorms were included in this data set. Each verified severe thunderstorm event during this period was investigated, as well as thunderstorms for which warnings were issued, but went unverified. Also included were any thunderstorms that approached (within 10 kg/m2) the critical VIL values from the OSF study. The highest VIL values recorded within 20 minutes before each severe weather event were used for this comparison, along with the highest VILs associated with each unverified warned event.
The greatest limitation to the database selected was the lack of complete echo top information. For those events where this information was missing, the average height between the last Base Reflectivity elevation slice with data for an individual storm and the next higher Base Reflectivity product without reflectivity returns were utilized to provide the closest estimate possible for echo height. Reflectivity cross-sections were also employed to provide an approximation of echo tops. If no reliable estimation of a storm's echo height was available, the event was discarded from the study.
A second limitation was a shortage of non-severe events within the database. Archive IV at NWSO Goodland is typically employed only when severe thunderstorms are considered possible. The database is therefore skewed toward severe thunderstorm events.
RESULTS OF THE COMPARISON
Section 5, Chapter C-72 of the Weather Service Operations manual (1987) supplied the following definitions that apply to this study:
|Event||-||A report of hail greater than or equal to 19.05 mm or 0.75 inch is recorded as an event.
|Warned Event||-||Any event (hail greater than or equal to 19.05 mm or 0.75 inch) that occurs within an area (county or portion of a county) for which a warning has been issued and within the valid time of the warning. For severe local storms, multiple reports occurring within 10 statute miles and 15 minutes of each other are recorded as one event.
|Areas Warned||-||Counties or portions of counties for which severe local storm warnings were issued.
|-||Counties or portions of counties for which severe local storm warnings were issued, but no event (hail greater than or equal to 19.05 mm or 0.75 inch) was reported.|
Of the 42 of thunderstorm events studied, 28 produced hail greater than or equal to 19.05 millimeters (0.75 inch). Using the VIL criteria for the occurrence of large hail as developed in Oklahoma as a basis for issuing a severe thunderstorm warning, and following the methodology of severe thunderstorm warning verification outlined above, the following was the result.
|Number of Warned Events||=||20|
|Total Number of Events||=||28|
|Number of Warned Areas Unverified||=||9|
|Number of Areas Warned||=||18|
The number of warned events exceeds the number of counties warned, as some storms produced more than one severe event in a particular county and, in a few cases, multiple severe thunderstorm warnings were issued for individual counties.
Scoring VIL using the warning verification criteria from Section 6, Chapter C-72 of the Weather Service Operations Manual (1987) produced the following:
|VIL Probability of Detection (POD)||=||.714|
|VIL False Alarm Ratio (FAR)||=||.500|
|VIL Critical Success Index (CSI)||=||.417|
To maintain consistency with the methodology used in the Oklahoma study, the VIL Density criteria for a 50 percent probability of large hail were employed as a basis for issuing a severe thunderstorm warning. Using VIL Density in this manner produced these values:
|Number of Warned Events||=||25|
|Number of Events||=||28|
|Number of Warned Areas Unverified||=||13|
|Number of Areas Warned||=||38|
Scoring VIL Density for a 50 percent probability of large hail, using the warning verification criteria from Section 6, Chapter C-72 of the Weather Service Operations Manual (1987) resulted in the following values:
|VIL Density (50%) Probability of Detection (POD)||=||.893|
|VIL Density (50%) False Alarm Ratio (FAR)||=||.342|
|VIL Density (50%) Critical Success Index (CSI)||=||.610|
A total of 18 thunderstorms met or exceeded the VIL Density criteria for a 90 percent probability of large hail (4.0 g/m3 or 1.22 kg/m2/kft) from the Tulsa study. Of these thunderstorms, only 12 produced reports of hail greater than or equal to 19.05 mm or 0.75 inch, resulting in a probability of large hail of only 67 percent.
A VIL Density correlating to a greater than or equal to 50 percent probability of large hail proved to be much more accurate in the identification of severe thunderstorm events than the VIL values as developed by OSF, yet suffered from over-warning on several non-severe thunderstorms. VIL values correlated with large hail, as developed for Oklahoma, proved to be less effective at identifying severe thunderstorms in the NWSO Goodland, KS CWA, and worse than VIL Density at distinguishing non-severe thunderstorms, based on VIL and VIL Density False Alarm Ratio (FAR) comparisons.
The failure of the VIL Density value for a 90 percent probability of large hail from the NWSO Tulsa, Oklahoma study to correlate to a similar value for the NWSO Goodland Kansas CWA may be due to several factors. Perhaps the most important of these is the smaller sampling of thunderstorms in the data set used in this study. The number of severe and non-severe thunderstorms used for this paper was approximately 25 percent of the number of thunderstorms investigated in the NWSO Tulsa study, so random error would have played a role. Lower population density over the NWSO Goodland CWA as compared to NWSO Tulsa's CWA may also have played an important role in reducing the expected accuracy of VIL Density at the theoretical 90 percent probability level as well. Despite an aggressive severe local storm verification program at NWSO Goodland, vast unpopulated regions limit the number of severe thunderstorm reports solicited from the public.
The results of this study, if substantiated by further research, indicate that VIL Density may prove to be a useful tool for the identification of severe thunderstorms. VIL Density provides improvements over VIL in the detection of severe thunderstorms. However, using VIL Density, as a "stand alone" severe weather indicator, may produce an unacceptable number of unverified warnings. Other factors, including storm environment and structure, need to be included in the warning decision process.
Amburn, S., and P. Wolf, 1995: VIL Density as a Hail Indicator. Presented at the Southern Region WSR-88D Conference, 6pp.
U.S. Dept. of Commerce, 1993: WSR-88D Operations Training Manual. NOAA, NWS, Operations Training Branch, Operational Support Facility, Norman, Oklahoma.
____________, 1987: National Weather Service Operations Manual, 1987: National watch/ warning verification program, NOAA, NWS, 19pp.