a. Synoptic conditions

The earliest evidence of the tornado on the ground was at 1840 UTC. Figure 1a shows an infrared satellite image with a 500-mb height and temperature analysis and Fig. 1b shows the 500-mb wind and absolute vorticity analysis from the 1800 UTC Rapid Update Cycle (RUC; Bleck and Benjamin 1993). An upper-level trough was located over the northern Rockies and extended south into the Great Basin. The coldest air was located at the base of the trough, over western Utah. The absolute vorticity analysis shows one short-wave trough axis extended from along the Montana–Idaho border southeast into southwest Wyoming, while another trough axis extended from the Oregon–Idaho–Nevada triple point into western Utah. Extensive clouds are seen ahead of the eastern short wave. This feature had passed through northern Utah during the previous night bringing precipitation to all of northwest Utah. The 700-mb RUC analysis of wind and temperature from 1800 UTC is shown in Fig. 2a. A closed circulation was centered over northern Idaho with the coldest air extending southwest into western Nevada. The 700-mb cold pocket was considerably west of the coldest air at 500 mb and there was actually warm advection taking place over northwest Utah, beneath the coldest air at 500 mb. At 300 mb (Fig. 2b) a closed low can be seen over Idaho with a 36 m s⁻¹ (70 kt) jet maximum on the front side of the trough over eastern Utah. The shear vectors in the 850–400-mb layer (Fig. 2c) indicate 15.4–18.0 m s¹ (30–35 kt) of vector shear in this layer near SLC. The 850–400-mb layer is used operationally in the Great Basin as a tool indicating whether supercell thunderstorms are likely in the same way that 0–6-km shear is used in many areas east of the Rockies (Weisman and Rotunno 2000).

The surface pressure trough extended westward across the Great Salt Lake (GSL) (Fig. 3). Moderate south winds of 5.1–10.3 m s⁻¹ (10–20 kt) can be seen at a few of the NWS's METAR sites south of the pressure trough. Temperatures for 1800 UTC (noon local time) were cooler than normal for this desert location, and dewpoint temperatures were higher than normal. This was a remnant effect from the nocturnal precipitation.

In summary, the synoptic-scale environment over northwest Utah at 1800 UTC included mid- and upper-tropospheric dynamical lift from the trough aloft, a surface pressure trough, moderate vertical wind shear, and a rapidly destabilizing air mass as cold air aloft moved over warm advection at lower levels.

b. Mesoscale conditions

The topography of northern Utah can be seen in Fig. 4. Salt Lake City is located at the northeast edge of a flat valley that is generally 1.3 km (4200 ft) MSL and is approximately 27 km in width. The Wasatch Mountains rise steeply to the east and a series of smaller mountain ranges are located to the west and southwest of the city. The mountains immediately southwest of SLC are the Oquirrh Mountains, which rise to over 3.05 km (10 000 ft) MSL. The GSL is located northwest of the city. This shallow, saline lake is nearly always cooler than the surrounding land on summer days.

Visible satellite imagery (Fig. 5) shows convection developing over the Oquirrh Mountains at 1730 UTC and then moving east over SLC between 1745 and 1830 UTC. The 1830 UTC visible image (Fig. 5c) shows a strong thunderstorm with a possible overshooting top to the south of SLC. This cell produced 3.81-cm (1.5 in.) diameter hail and had a weak mesocyclone at midlevels, but was not responsible for the tornado. The tornadic cell is difficult, if not impossible to identify in satellite imagery (Fig. 5d).

The surface observations and analysis in Fig. 3 represent only METAR data. The sea level pressure trough in this figure appears to be centered over the GSL. An extensive network of automated surface observations exists across northern Utah (Slemmer 1998). This network is a collection of cooperating public agencies (state and federal) and private entities and is referred to as MesoWest. The stations vary in temporal resolution from every 5 min to every 3 h, with the majority of sites reporting every 15 min to hourly. Data from MesoWest at 1800 UTC are displayed in Fig. 6. The data in the plot are from observations within the 15 min prior to the time on the bottom of the plot. At 1800 UTC the stations along the south end of the lake were all showing northerly winds with temperatures in the upper 60°s to lower 70°s (Fahrenheit), while farther south, in the Salt Lake Valley and the Tooele Valley, the winds were from the south and the temperatures were in the mid 70°s. The convergence zone implied by these observations is characteristic of the lake-breeze front that moves inland away from the lake nearly every summer day around the GSL (Astling et al. 1995). The convergence zone could also be seen at 1815 and 1830 UTC (not shown). There was little or no movement in the convergence zone at those times. Also shown in Fig. 6 is the location of a convergence–shear boundary as seen in TDWR radar data discussed later. The radar-derived boundary lines up fairly well with the MesoWest data where the boundary is oriented northeast–southwest, although there are not quite enough surface observations northwest of the boundary for a confident boundary placement at the surface. Along the western end of the radar-derived boundary, there are surface observations of southerly winds on the north side of the boundary indicating the boundary tilted with height, or that there was actually more than one boundary, although the radar data showed only one boundary.

Although the METAR data showed a synoptic-scale surface trough over the GSL, the MesoWest data indicated that a mesoscale convergence boundary was actually located just south of the lake. The origins of the boundary are uncertain. It was present earlier in the morning before the onset of any convection (not shown). The boundary was either an early manifestation of the diurnal lake breeze, the remnant from the nocturnal precipitation, or the synoptic-scale front itself. By 1800 UTC, convection over the Oquirrh Mountains and the southern end of the GSL had begun to interact with this boundary. Regardless of the origin of the boundary, tornadogenesis took place in the vicinity of the boundary around 1830 UTC, west-southwest of the immediate downtown area.

The 1200 UTC sounding from SLC is shown in Fig. 7a. Also shown is the 1800 UTC analyzed sounding (Fig. 7b) from the Utah Advanced Regional Prediction System (ARPS) Data Analysis System (ADAS; Ciliberti et al. 1999; Brewster et al. 1995). The Utah ADAS is a three-dimensional mesoscale analysis system run in real time at the University of Utah. The domain is centered on SLC with 1-km horizontal resolution and 33 vertical levels. The RUC2 is used as the initial background field. A comparison of the 1200 and 1800 UTC soundings reveals a 2°–3°C cooling of 500-mb temperatures, resulting in steeper lapse rates by midday. The winds have backed in the lower levels and veered aloft, indicating low-level warm advection. Finally, the magnitude of the vector vertical shear, the vector difference between these two surfaces, remained near 18.0 m s⁻¹ (35 kt) between 850 and 400 mb.

Figure 8 shows the 1800 UTC convective available potential energy (CAPE) analysis from the Utah ADAS. The CAPE in the area just southeast of the GSL along the aforementioned convergence boundary was approximately 1500 J kg⁻¹, based on lifting the analyzed surface parcel. The corresponding lifted index (not shown) was −8.

In summary, the mesoscale observations and analyses show a convergence boundary was just south of the GSL in the vicinity where subsequent tornadogenesis would occur. The air mass destabilized significantly in the 6 h between 1200 and 1800 UTC as a result of low-level warm advection and strong cooling at 500 mb. By 1800 UTC convection initiated on the nearby terrain southwest of SLC, and joined the instability and stationary convergence boundary as key ingredients for the tornadic storm.

c. Tornado damage path

The damage path of the tornado is shown in Fig. 9 and a photo of the tornado is shown in Fig. 10. The total pathlength was 6.84 km (4.25 mi). The average width of the path was 91.44 m (100 yd). The damage path just northeast of downtown was more than 400 m (0.25 mi) in width. The tornado caused damage along the path continuously from 1841 to 1855 UTC. The tornado path climbed from 1.29 km (4225 ft) MSL at the start of the path to 1.62 km (5320 ft) MSL at the end of the path, including a descent into a canyon and back up the other side of the canyon approximately midway through the path.

The tornado resulted in one fatality, 80 injuries, and damage or destruction of approximately 300 buildings. Some of the structural damage occurred to very large buildings in the downtown area. Thirty-four houses in the residential areas were left uninhabitable, and approximately 800 trees were damaged or destroyed. Damage estimates to the residential areas were approximately $170 million.

Real-time and time-lapse videotape of the tornado taken from east of the path showed the tornado began initially as a somewhat disorganized circulation disconnected from the cloud base. The tornado appeared to intensify very rapidly a few minutes after genesis, and the debris cloud and funnel cloud merged. Eventually a lowered wall cloud was visible in the video. As the tornado moved through the residential areas, the debris cloud became less visible. The tornado became very thin toward the end of the damage path, even though some of the worst damage occurred in this area.

a. Radar characteristics and beam geometry

The SLC tornado was observed by two Doppler radars. Figure 4 shows the locations of the KMTX WSR-88D and the TDWR with respect to the tornado damage path. The TDWR is located 22.5 km (12 n mi) north of the damage path at the same elevation as SLC, while KMTX is located 76 km (41 n mi) to the northwest and 0.7 km (2300 ft) above SLC. The beam geometry for the two radars with respect to the tornado is shown in Fig. 11 and the characteristics of the two radars are shown in Table 1.

The TDWR has a much smaller beamwidth and was much closer to the tornado than KMTX. As seen in Fig. 11, there are four elevation angles from the TDWR that intersected the tornado below the midpoint of the lowest angle from KMTX. The beam diameter at the point of intersection with the tornado is approximately 152 m (500 ft) for the TDWR compared to 1310 m (4300 ft) for the lowest KMTX beam. The midpoint of the TDWR 0.3° elevation angle is approximately 152 m (500 ft) above ground level (AGL), while the midpoint of the lowest KMTX beam is approximately 1524 m (5000 ft) AGL. The TDWR scan strategy is primarily made up of sectors near and over the SLC airport. Fortuitously, the tornado was captured within this sector until that time when it moved into the high terrain northeast of downtown SLC near the end of the event. Also of note is that with the TDWR's 5-cm wavelength somewhat greater attenuation can be expected when compared to the higher-powered KMTX with its 10-cm wavelength. This might be expected to change the reflectivity values somewhat compared to KMTX, but the signal attenuation does not impact velocity estimates as long as there is some returned signal (Johnson and Brandes 1987).

b. Analysis of the TDWR data

The lowest five tilts of velocity data and one reflectivity tilt from the volume scan that began at 1824 UTC are shown in Fig. 12. The convergence–shear boundary can be seen in the 0.3° tilt (Fig. 12a), but there was considerable ground clutter causing blank data areas. The boundary is clearly seen in the 1.0 and 2.0° tilts (Figs. 12b and 12c), but is not visible in the 4.1° or 6.3° tilts (Figs. 12d and 12e). Thus, the vertical depth of the boundary at 1824 UTC was between 1.03 and 1.77 km (3400 and 5800 ft) AGL. The reflectivity data aloft at 1824 UTC (Fig. 12f) show a region of disorganized echoes with the eastern edge of the higher returns aligned with this convergence–shear boundary. There were no echoes at low levels in the vicinity of the boundary at this time (not shown).

Five minutes later, at 1829 UTC, the convergence–shear boundary has deepened and the implied circulation has strengthened. The lowest six tilts of velocity data are shown in Fig. 13; the same five as in the previous figure plus the 8.3° tilt are shown. The outbound velocity has increased on the lowest three tilts while the inbound velocity has remained unchanged, indicating a strengthening of the low-level circulation (Figs. 13a–c). The convergence zone was much better defined in the 2.0° tilt (Fig. 13c) than 5 min earlier. The circulation was visible in the 4.1° and 6.3° tilts (Figs. 13d,e), but not at all in the 8.3° tilt (Fig. 13f). The circulation at 1829 UTC extended upward to between 2.62 and 3.14 km (8600 and 10 300 ft) AGL. Arrows in the figure point to the incipient vortex.

By 1834 UTC the convergence boundary had continued to build upward and the circulation had strengthened further. The same lowest six tilts of velocity data are shown in Fig. 14, along with velocity data from two tilts higher up in the atmosphere. The outbound velocities continued to increase at low levels (Figs. 14a–d) between 1829 and 1834 UTC while inbound velocities remained relatively unchanged, indicating the convergence zone had strengthened further. The circulation was now visible on the 8.3° tilt (Fig. 14f) at 1834 UTC. The circulation was centered near the reference point in the lowest elevation angles and appears to tilt slightly to the northwest with height. The circulation had now reached a height of at least 3.14 km (10 300 ft) AGL. Velocity data from the 17.4° and 20.7° tilts are shown in Figs. 14g and 14h. Divergence can be seen near the reference point, which was directly above the circulation center on the lowest tilts. This was the first volume scan in which the divergence signature was clearly visible and indicates the presence of an updraft (Burgess and Lemon 1990) over the lower-level circulation.

At 1838 UTC, just 3 min before tornado damage was first noted on the ground, the velocity data showed the strength of the convergence and circulation continued to increase (Fig. 15). The outbound velocities near the circulation center were visibly stronger at 2.0°, 4.1°, and 6.3° tilts (Figs. 15c–e) than on the previous volume scan. In addition, the outbound and inbound velocity maxima on the 1.0° tilt (Fig. 15b) were on adjacent azimuths (a gate-to-gate velocity signature). Ground clutter contamination on the 0.3° tilt prevented the inbound velocity data from being seen (Fig. 15a), so it is not clear if the vortex existed near the ground. The storm updraft had increased in depth and intensity as the storm-top divergence was no longer visible at 17.4° (Fig. 15g), but was much stronger at 20.7° (Fig. 15h). This updraft was located directly above the low-level vortex, suggesting the updraft may have been enhancing the circulation's vertical vorticity near the surface through stretching.

In comparing the low-level reflectivity data at 1838 UTC with the corresponding velocity data (Fig. 16) there was nothing remarkable about the reflectivity data associated with the incipient tornadic signature (manually assessed, not via algorithm) seen just north of the circular reference point in Fig. 16b. However, by 1843 UTC (Fig. 17) a well-defined hook echo and bounded weak echo region (BWER) were present in the reflectivity data and collocated with the tornado signature in the velocity data. The hook echo was present at 0.46 and 1.77 km (1500 and 5800 ft) AGL (Figs. 17a,b). At 2.62 km (8600 ft) AGL (Fig. 17c) the hook had completely encircled a region of weak reflectivity, and this weak echo region was capped by high reflectivity values at 5.7 km (18 700 ft) AGL (Fig. 17d). The BWER was in the same location as the tornado signature seen in the velocity data at 0.46 and 2.62 km (1500 and 8600 ft) AGL (Figs. 17e and 17f). Videotape of the event at this time showed a wall cloud at cloud base. These structural characteristics, a hook echo and BWER, are commonly associated with classic supercell-type tornadoes; although, Roberts and Wilson (1995) noted similar reflectivity structures with tornadoes that caused F2 damage in eastern Colorado that were not associated with supercell thunderstorms.

It is worth noting that the rather ill-defined reflectivity pattern at 1838 UTC is in sharp contrast to nondescending tornadoes that have been documented along the leading edge of squall lines and bow echoes. In these cases (e.g., Carbone 1983; Trapp et al. 1999) tornadogenesis takes place in a region of strong reflectivity gradient.

A time–height diagram of maximum differential velocity from 1822 to 1855 UTC is presented in Fig. 18 and summarizes the temporal and vertical evolution of the event. The values represent the difference between the maximum inbound and outbound velocity associated with the circulation that became the tornado. The bottom numbers are velocity differences (kt) from adjacent azimuths at constant range, commonly referred to as gate-to-gate delta υ (GTG). The other numbers are the maximum differential velocity in the overall circulation, again at the same range, but are not on adjacent azimuths. The delta-υ calculations that were not GTG were from velocity maxima that were within 5 km or less of each other. Shallow rotational convergence with differential velocity values of 7.7–15.4 m s⁻¹ (15–30 kt) are seen up through 1830 UTC. From 1830 to 1840 UTC the rotation deepened and strengthened, although the maximum GTG values remained less than 15.4 m s⁻¹ (30 kt). After 1840 UTC, which coincides with the visual observation of the tornado, the rotational and GTG values increased dramatically with a maximum GTG of 44.2 m s⁻¹ (86 kt) noted at 1847 UTC. After 1847 UTC, the tornado was no longer located within the TDWR sector scans and the data shown are from the three full 360° scans that are part of the volume coverage pattern. After 1855 UTC the tornado was no longer visible and damage was not reported as the remnants of the storm moved into the mountains just northeast of SLC.

The maximum vertical vorticity can be estimated from the maximum azimuthal Doppler velocity shear by the method described in Roberts and Wilson (1995). The vorticity is estimated by dividing the maximum azimuthal velocity difference in the rotation by the distance between these extremes. Assuming cylindrical symmetry of the rotation, the vorticity estimate is equal to twice the azimuthal shear. In this case the maximum vertical vorticity based on the TDWR data was 215 × 10⁻³ s⁻¹. This compares to maximum vertical vorticity values associated with the three tornadoes studied by Roberts and Wilson that ranged from 200 to 300 × 10⁻³ s⁻¹.

The tornadogenesis process was captured in considerable detail by the TDWR. In the 17 min prior to the formation of the tornado, the velocity data from the TDWR clearly showed a circulation forming along a near-surface convergence–shear zone, gradually building upward and intensifying. The intensification and vertical development of the circulation were coincident with increasing divergence aloft directly over the circulation. This circulation eventually contracted in scale and strengthened to form the tornado. Finally, after the tornado had formed, the TDWR reflectivity data showed a well-defined hook echo and a BWER.

a. Full-resolution WSR-88D velocity data analysis

The full-resolution WSR-88D velocity data has 0.25-km (0.13 n mi) range gates. These full-resolution data are archived (Crum and Alberty 1993; Crum 1995) and can be played back in non–real time using the WSR-88D Algorithm Testing and Display System (NSSL 1997). In the next subsection, a description of the different velocity products that are available operationally will be discussed. For this discussion, the full-resolution data are shown.

As seen in Fig. 11, only the 0.5° tilt from the KMTX radar sampled below 3.05 km (10 000 ft) AGL over SLC. The 0.5° base velocity data from the 1831 UTC volume scan (10 min before the tornado) are shown in Fig. 19a. The low-level convergence zone can be seen with an area of outbound velocities adjacent to an area of inbound velocities to the west of SLC. Embedded within the inbound area are some strong outbound velocities. This feature results in a number of very strong anticyclonic GTG values. This noisy field is often seen at this location, which is where two interstate highways converge at the Salt Lake City International Airport. Slemmer (1998) noted the persistence of this signature in situations of low reflectivity in a radar climatology study of snowstorms over SLC and attributed it to contamination from airplanes and vehicle traffic. There was one moderate GTG value in the cyclonic sense along the convergence boundary, but the validity of this single value is questionable given the overall contamination of the data at this time. This feature may represent another circulation that did not become tornadic, and time-lapse video from the University of Utah Meteorology Department did show other circulations in the cloud base in the period before tornadogenesis.

At 1836 UTC the outbound velocity values that were previously embedded in the inbound region were no longer present (Fig. 19b). Higher reflectivity values near the airport suggest that the signal to noise ratio had increased, resulting in a more reliable velocity estimate in this region. The overall strength of rotational convergence had increased along the boundary. The maximum GTG value was 28.8 m s⁻¹ (56 kt), which is actually higher than the maximum GTG value from the TDWR data at this time (Fig. 18). One possible explanation is the circulation in the TDWR data is much wider than the beamwidth and, hence, the maximum inbound and outbound velocities are not along adjacent azimuths. The maximum delta v from the TDWR at this time was 26.75 m s⁻¹ (52 kt). The larger beamwidth of the WSR-88D made it impossible to detect the weaker velocity values near the center of the circulation at this time. Another possible explanation is that the strong, inbound velocity value in the KMTX data was the result of noisy returns as was seen in the previous volume scan (Fig. 19a). If these data were disregarded, then the maximum GTG value at this time was 15.4 m s⁻¹ (30 kt).

The velocity data from 1841 and 1846 UTC are shown in Figs. 19c and 19d. The tornado had formed by these times. The locations of the largest GTG values for the tornadic circulation are indicated in the figures. Comparison with the TDWR velocity data at these times (Fig. 15 and Figs. 17e and 17f) shows that the overall circulation feature is captured, but with less detail and a more disorganized appearance than in the TDWR data. A time–height diagram of the delta v from KMTX is shown in Fig. 20. The vertical and temporal evolution seen by KMTX was similar to that seen by TDWR (Fig. 18) in that the circulation either builds upward or appears at multiple levels simultaneously, but with weaker maximum rotational and GTG values. As noted, the GTG value at the lowest tilt at 1836 UTC is questionable. Although the KMTX lacks the detail of the TDWR data, it samples the tornadic circulation well, perhaps better than might be expected considering the elevation above the valley, range from the storm, and the sampling volume of the WSR-88D. The hook echo and BWER seen in the TDWR reflectivity data were not visible at all from KMTX (not shown).

b. Operational WSR-88D velocity products

The data shown in the previous section represent the raw base velocity resolution output by the WSR-88D at 0.25-km (0.13 n mi) range by 1° of azimuth. However, the real-time base velocity products limit the highest resolutions to within a 60-km (32 n mi) radius from the radar. The finest real-time product resolution degrades to 0.50 km (0.27 n mi) for ranges between 61 km (33 n mi) and 115 km (62 n mi) from the radar. Beyond the 115-km (62 n mi) range, the maximum resolution degrades to 1.0 km (0.52 n mi). Each 0.50-km (1.0 km) range bin is a display of the nearest of the two (four) 0.25-km range bins that are geographically encompassed by the larger bin. The SLC tornado was approximately 76 km (41 n mi) from the radar, and, thus, only every other bin was displayed in the vicinity of the tornado.

The base velocity products from the 0.5° tilt at 1836, 1841, and 1846 UTC are shown in Fig. 21 with the location of the maximum GTG value indicated. Comparison with Figs. 19b–d reveal a much degraded resolution. In addition, the real-time products do not allow for interrogation of the actual data value, but rather each color represents a range of velocities. At 1836 UTC, the full-resolution data showed a maximum GTG value of 28.8 or 15.4 m s⁻¹ (56 or 30 kt) depending on whether one believes the data to be contaminated or not. If the maximum inbound and outbound values from the range depicted in the real-time product color table are used, then an estimate of the maximum, or “worst case,” GTG value can be obtained. At 1836 UTC, the worst case GTG value was 21.1 m s⁻¹ (41 kt). At 1841 UTC, the full-resolution GTG value was 25.7 m s⁻¹ (50 kt), while the real-time product maximum worst case GTG was 31.4 m s⁻¹ (61 kt). Using the midrange of the color bars as is done by some operational forecasters, the GTG value was 26.0 m s⁻¹ (51 kt). At 1846 UTC, at the time of F2 intensity, the full-resolution GTG value was 29.8 m s⁻¹ (58 kt), while the real-time product maximum GTG, using the extremes from the color range, is 38.6 m s⁻¹ (75 kt) (midrange GTG 32.4 m s⁻¹). The locations of the maximum GTG values are the same in both the degraded and full-resolution products.

Another real-time velocity product created by the WSR-88D is the storm-relative mean radial velocity (SRM). This product is used extensively in real-time operations because it subtracts out a mean storm motion vector from all data points and, at times, can make it much easier to see rotational features. The mean vector in use on KMTX on this date was 275° at 6.2 m s⁻¹ (12 kt). The spatial resolution of the SRM is 0.54 nm, regardless of range from the radar. The SRM is created by taking the maximum absolute value of the velocity from consecutive, nonoverlapping groups of four range bins, which is in contrast to the base velocity product that uses the first value from the groups of two (0.27 n mi) or four (0.54 n mi) range bins. For example, the SRM may actually display an inbound value where the base product shows an outbound value, if the group of four range bins contained both inbound and outbound values, and the range bin closest to the radar was outbound, but one of the other four bins was inbound with a larger absolute value. This situation appears to have taken place in the event studied here.

Figure 22 shows the SRM product for 1836 UTC with the location of the largest GTG value in the full-resolution data and the largest GTG value in the SRM indicated by arrows. Comparing the full-resolution base velocity at 1836 UTC (Fig. 19b) with the SRM at the same time (Fig. 22) shows that the strongest GTG values seen in the base velocity were no longer present in the SRM, and only inbound velocities were seen. Table 2 shows the velocity values in the individual range bins on the two adjacent radials at the location of the strongest GTG value in the base data. When the groups of four gates were evaluated for the SRM, the absolute value of the outbound gate was smaller than the inbound gates, and the SRM showed only inbound velocity to the west of the strongest inbound velocity, instead of weak outbound. The worst case GTG value at this location in the SRM was only 10.3 m s⁻¹ (20 kt). The midrange GTG at this same location is 5.1 m s⁻¹ (10 kt). A GTG value of 26.8 m s⁻¹ (52 kt) can be found in the SRM at 1836 UTC using the maximum values of the color ranges, but it is not in the same location as the maximum value in either the full-resolution or real-time velocity data. It is located approximately 16 range bins to the northwest of the incipient tornado circulation in the full-resolution data, and may be associated with the aforementioned other circulation feature seen at cloud base in the time-lapse video. The SRM products at 1841 and 1846 UTC (not shown) were also different in appearance than the base velocity data, but the location of the maximum GTG value was the same.

c. Tornado detection and mesocyclone algorithm output

Although the real-time velocity products produced by the WSR-88D have degraded presentations of the full-resolution data as a function of range, the TDA and Mesocyclone Algorithm both use the full-resolution velocity data and produce real-time output for operational use. The TDA was recently updated with a new version described by Mitchell et al. (1998), and this was the algorithm running on the KMTX system in August 1999. The TDA searches the velocity data using a variety of GTG value thresholds and vertical continuity criteria. Unlike previous versions of the TDA, the new TDA is no longer coupled to the Mesocyclone Algorithm. The presence of a mesocyclone is no longer a prerequisite for an algorithm detection of a tornado vortex signature. The basic methodology of the Mesocyclone Algorithm in use during this event is described by Zrnic et al. (1985) and a description of strengths and weaknesses of the algorithm can be found in Stumpf et al. (1998).

A TDA detection was triggered on the volume scan that began at 1836 UTC. Examination of Fig. 19b and Fig. 20 indicates that the vertical continuity criteria for the TDA was met, but the only level where the GTG value was large enough to meet the algorithm criterion was on the 0.5° tilt using the somewhat questionable gates in the vicinity of the SLC airport. The TDA triggered again on the 1841 and 1846 UTC volume scans, but not at 1851 UTC or later. The Mesocyclone Algorithm triggered only on the 1841 UTC volume scan.

In this case, the TDA would appear to have provided useful guidance to the operational forecaster. However, there are some limitations and issues with the utility of the algorithm output for this event. The TDA detection from the volume scan beginning at 1836 UTC was generated in real time and delivered to the forecaster at the end of the volume scan at 1840 UTC. The tornado was first observed on the ground at this time, so the algorithm did not provide any lead time for a tornado warning. Another operational issue is that upon receipt of the algorithm notification, the forecaster would have likely examined the velocity data for confirmation of the tornado vortex signature. Corroboration would have been sought in this case for at least three reasons: 1) a preexisting supercell thunderstorm was not already present in the area of the TDA detection; 2) false TDA detections from the KMTX radar are not uncommon and have occurred on a number of occasions, including due to the passing of a freight train in the desert just west [but 0.7 km (2300 ft) below] the radar site; and 3) the climatology of tornadoes in northern Utah might justify forecaster skepticism toward the algorithm output. Upon inspection of the 1836 UTC SRM (Fig. 22), the forecaster would have been unable to identify a region of strong GTG values for the reasons discussed in the previous section. Examination of the real-time base velocity product (Fig. 21a) may not have led to confidence in the TDA detection either, because, even if the forecaster identified the incipient tornado in the blocky pattern, the maximum GTG value assuming a worst case from the color table velocity ranges only yielded 21.1 m s⁻¹ (41 kt), which is below the threshold used by the algorithm. The algorithm threshold was also not met on higher tilts in the real-time products (not shown). The only way the forecaster could have confirmed the validity of the algorithm would have been by examination of the full-resolution velocity data, which is presently not a real-time operational product produced by the WSR-88D. A severe thunderstorm warning was issued during this event at 1846 UTC that covered a number of storms in the area, including the one that became tornadic, but no tornado warning was ever issued for this event.