Performance of The Goodland WSR-88D During a Heavy Rainfall Event in Northwest Kansas
National Weather Service Office
The purpose of this paper is to detail the performance of the Goodland (GLD) WSR-88D during a heavy rainfall event that occurred at Oberlin in northwest Kansas on the early morning hours of July 15, 1994. Figure 1 is a background map of the area under study. The primary focus will be on the accuracy of the rainfall estimate of the Storm Total Precipitation (STP) product despite the presence of certain factors that could have made the product less accurate. The Vertically Integrated Liquid (VIL), high level (33,000-60,000 ft from MSL) Layered Composite Reflectivity Maximum (LRM), and Base Reflectivity products will also be examined.
Figure 1. Area where storm occurred.
2. THE EVENT AND OBSERVATIONS
Thunderstorms developed over northwest Kansas and extreme southwest Nebraska in the presence of the middle and upper tropospheric winds blowing from the northwest. These storms may have been the result of a vorticity maximum and its associated shortwave trough in the geopotential height field causing upward vertical motion due to vorticity advection increasing with height. The air mass over the region was also very moist. The observations from Goodland, Kansas and McCook, Nebraska indicated dew points from the upper 50s to the lower 60s.
The thunderstorm that produced nearly 3 inches/2.78/ of rain at Oberlin, Kansas, in roughly one hour and 30 minutes, resulted from a merger of two cells over extreme southwest Red Willow County of Nebraska and extreme northeast Rawlins County of Kansas (Figure 2). The VIL, high level LRM, and Base Reflectivity will be used to describe the sequence of events during this time. Based on their output and certain other factors, the accuracy of the VIL and the precipitation estimates were questioned.
Rain began in Oberlin around 0830 UTC (Figure 2). The thunderstorm that resulted from the merger then began to intensify, as shown by the base reflectivity as it moved closer to Oberlin. However, during this time, the VIL, from scan to scan, oscillated between high and low values.
Figure 2. The Base Reflectivity product from the 08:29 UTC volume scan on 15 July 1994.
The Volume Coverage Pattern (VCP) may have affected the VIL output. The VCP during this event was 21. When a storm is within 60 nm of the radar, the VIL will fluctuate more in VCP 21 than in VCP 11 since VCP 11 has fewer gaps than 21 (Operational Support Facility 1993). VCP 11 scan strategy involves more elevation angles than a strategy employed by VCP 21 (Figure 3). This type of fluctuation behavior was described by Mahoney (1987) in a study of VIL values from a single storm (Figure 4). The distance of the storm from the GLD WSR-88D ranged from 45 nm to 60 nm.
Figure 3. Illustration of volume coverage patterns 11 and 21.
Figure 4. Fluctuations of VIL values as a function of range for VCP 11 and VCP 21 respectively.
The STP showed a swath of very heavy rainfall heading toward Oberlin. Since the VIL estimate may not have been accurate, the accuracy of the STP was in doubt as well. However, just because the VIL was "bad" did not necessarily mean the precipitation estimate was also poor. The input into the precipitation product includes only the base reflectivity from the four lowest elevation angles. The VIL uses reflectivity data from the entire volume scan (Federal Meteorological Handbook No. 11 1991).
Another factor may have been the storm speed. The movement of the storm as shown by the GLD WSR-88D was around 35 mph. This contributed to the disregard for the extremely high precipitation estimates. The faster the storm, the less time it spends over a certain location to produce substantial rainfall.
At 0858 UTC, according to the VIL, high level LRM, and Base Reflectivity products, the heavy rainfall moved into Oberlin (Figures 5-7). The reflectivity stayed very high over the town until shortly after 0922 UTC (Figure 8). At this time, roughly 0858 UTC to about 0922 UTC, another occurrence happened which indicated that the VIL was not representative, and the precipitation was being greatly overestimated.
Figure 5. The high level LRM product from the 08:58 UTC volume scan on 15 July 1994.
Figure 6. The VIL product from the 08:58 UTC volume scan on 15 July 1994.
Figure 7. The Base Reflectivity product from the 08:58 UTC volume scan on 15 July 1994.
Figure 8. The Base Reflectivity product from the 09:22 UTC volume scan on 15 July 1994.
The Decatur County Sheriff reported half inch diameter hail at Oberlin during this time. One report indicated hail from 0900 UTC to 0915 UTC while another report at roughly the same time stated that hail occurred for 20 minutes. The VIL as displayed in Figure 6 was high at the beginning of this period of hail. The assumption was then made that hail could have been the cause for the high VILs and high precipitation estimates to the northwest of Oberlin. With the previous factors mentioned, it was concluded that the STP output was not representative.
Another inconsistency developed between the products during the time of the hail. While the base reflectivity was relatively very high, the VIL and high-level LRM products from 0910 UTC to the end of the event had low values over Oberlin. Although the two products did not indicate the storm was very strong, the base reflectivity showed that heavy rainfall did not end until about 0945 UTC (Figure 9). Light rainfall continued to 1000 UTC that is when the storm left the area.
At 1015 UTC a report was received which confirmed the high rainfall amount that the STP had indicated. The Decatur County Sheriff reported 2.78 inches had fallen in an hour and a half. According to the GLD WSR-88D data, the time of the rainfall was from 0830 UTC to 1000 UTC. However, the base reflectivity indicated that most of the rain occurred from around 0900 UTC to just before 0945 UTC.
Figure 9. The Base Reflectivity product from the 09:45 UTC volume scan on 15 July 1994.
The STP total had Oberlin barely inside the 4-inch area with the 3 to 4 inch area just to the west (Figure 10). This would mean that the STP was only a little over an inch above the actual amount. The STP also proved to be very accurate on either side of the path of heaviest rainfall. McCook received 1.61 inches and an observer 16 miles northeast of Hoxie reported 1.6 inches. This corresponded very well to the STP estimate (Figure 11). Since the thunderstorm that hit Oberlin was between McCook and the GLD WSR-88D, it was even more remarkable that the estimate for McCook was so accurate. Despite being a fast-moving storm, having fluctuating VIL values due to being in VCP 21, and having half inch diameter hail fall from the thunderstorm, the Storm Total Precipitation product's estimate was within an inch of the actual amount.
Figure 10. Blow-up of STP product from the 11:31 UTC volume scan on 15 July 1994.
Figure 11. STP product from the 11:31 UTC volume scan on 15 July 1994.
The author would like to thank Mr. John Kwiatkowski (SOO) and Mr. Scott Mentzer (MIC) for their review and valuable suggestions as to the organization of this paper.
Mahoney, E.A., 1987: Limitations of the Vertically Integrated Liquid Water Algorithm during the NEXRAD Era. Master's Thesis, University of Oklahoma, Norman, OK.
NOAA, DOC, OFC, 1991: Federal Meteorological Handbook No. 11 - Doppler Radar Meteorological Observations. Part C WSR-88D Products and Algorithms. Government Printing Office (GSA), 34-35.
Operational Support Facility, 1993: Student Guide for WSR-88D Training Course. Topic 8 Lesson 2 Vertically Integrated Liquid. Operations Training Branch, Operational Support Facility, U.S. Department of Commerce, NOAA, National Weather Service, 8-2-1 to 8-2-5.