1. Introduction
Since tornadoes are a significant hazard to lives and property, it is important to understand why they form so that they might be more accurately predicted. It is also important to know the three-dimensional distribution of their horizontal and vertical wind components, so that the nature of the damage they inflict might be better understood. The primary focus of this paper is on the structure of the wind field.
While laboratory models and numerical simulations have been used to estimate the horizontal and vertical components of the three-dimensional wind field in tornadoes, both are subject to serious limitations. In the case of the former, the behavior of vortices and how they interact with the ground has been studied, but the parent storm has not been accounted for and it is difficult to account for flows that are azimuthally asymmetric, as they are in nature. In the case of the latter, efforts to do controlled experiments including both the tornado and the parent storm have been a substantial challenge, owing to the high spatial resolution required and limited computing resources.
Making measurements in real tornadoes is a substantial challenge in its own right, owing to the small size of tornadoes, their very low frequency of occurrence, and the dangers and difficulties in getting instruments close to them. Using mobile Doppler radars has proven to be an effective and safe method for determining the structure of tornadoes (e.g., Wurman et al. 1996; Wurman and Gill 2000; Wurman 2002; Bluestein et al. 2003a, b; Bluestein et al. 2004b; Lee and Wurman 2005; Alexander and Wurman 2005; Wurman and Alexander 2005; Tanamachi et al. 2007). To resolve the structure of a tornado, the radar must be able to resolve volumes with each dimension on the order of tens of meters, since most tornadoes are on the order of hundreds of meters. Mobile X- (∼3-cm wavelength) and W-band (∼3-mm wavelength) radars have been used to attain high enough spatial resolution to resolve tornado structure (e.g., Wurman et al. 1997; Bluestein and Pazmany 2000). The two main factors that determine the spatial resolution in the cross-beam direction are the range from the tornado and the antenna beamwidth. Operating at a range of 3–5 km seems to be a good compromise for being able to see the tornado and for safety (Bluestein 1999); positioning the radar within 1.5 km of a tornado can be dangerous, while positioning the radar beyond 6 km, when the visibility is low, may be too far away to see the tornado visually. The current X-band radars have antennas whose half-power beamwidth is ∼1°, while the current W-band radar has an antenna whose half-power beamwidth is 0.18°. Thus, the spatial resolution of the X-band (W band) system is ∼50 m (∼10 m) at 3-km range. While the W-band radar system has the higher spatial resolution, it suffers from severe attenuation when large hydrometeors surround the tornado. It is therefore useful to probe tornadoes with both radar systems; the X-band system can resolve both the tornado and its parent storm, while the W-band system can resolve the tornado, but with ∼5 times the spatial resolution in the directions normal to the beam.
The main issues concerning tornado structure include the following: 1) the nature of its vertical circulation as a function of swirl ratio (Davies-Jones et al. 2001); 2) the variation of horizontal wind speeds with height (Lewellen et al. 2000) and, especially, the height above ground and radius of the strongest horizontal wind speeds; and 3) the reflectivity structure, including the “weak-echo hole” (e.g., Wakimoto and Martner 1992; Bluestein et al. 2004b; Dowell et al. 2005). The primary purpose of this paper is to address some of these issues using mobile Doppler radar datasets, photographs, and videos acquired on 12 May 2004 in south-central Kansas. On this date, a supercell developed from ordinary convective cells that had formed near the intersection of a dryline and outflow boundary (see the appendix for a brief discussion of the synoptic environment) around 1730 CDT. Numerous tornadoes, some of which were probed by mobile X- and W-band Doppler radars, were observed beginning ∼1.5 h later (Fig. 1). This paper includes analyses of data collected by the X- and W-band radars designed and built at the University of Massachusetts—Amherst (UMass) and operated by the University of Oklahoma (OU). Only reflectivity and Doppler velocity will be considered here. Dual-polarization reflectivity data collected by the UMass X-band radar (Bluestein et al. 2007) and Doppler radar data collected by the X-band Doppler on Wheels (DOW; Kosiba et al. 2005) are discussed elsewhere. Also detailed elsewhere are thermal observations from an infrared camera located at the radar sites (Tanamachi et al. 2006).
The unique aspects of the UMass radar datasets are as follows: 1) The quality of this (12 May 2004) W-band radar dataset is superior to that of the best previous W-band dataset collected in 2002 near Happy, Texas (Bluestein et al. 2004b). Both sector scans near the ground and range-height indicators (RHIs) were obtained at close range in several tornadoes, along with boresighted video that allowed us to identify the portions of the tornado that were scanned. 2) Both X- and W-band radar data were collected nearly simultaneously from approximately the same locations, so that an intercomparison between the two is possible in some instances. In addition, it is possible to link features seen in the W-band dataset on the tornado scale with features seen in the X-band dataset on the storm scale.
The following section (section 2) contains a description of the radar systems and the nature of and specific problems involved with the data processing. Section 3 contains a detailed exposition of the analyses of the Doppler radar data collected in several tornadoes. A summary and discussion of the results are given in section 4.
2. Description of the radar system and data processing
The W- and X-band, mobile Doppler radars were built at the Microwave Remote Sensing Laboratory (MIRSL) at UMass. Each radar system is mounted on a truck. The antennas and their pedestals are mounted on the rear of each truck; a computer located inside the truck controls the operation of the radars. The radar operator sits behind the navigator, who is in the front seat on the passenger side. Time series of data are collected and stored on the computer. The data are postprocessed at MIRSL and made available in a format compatible with the SOLO editing and display software from the National Center for Atmospheric Research (NCAR; Oye et al. 1995).
a. W-band radar
Details about this radar are found in Bluestein and Pazmany (2000) and elsewhere. The unique aspect of this radar is that its half-power beamwidth is only 0.18°, which allows for very fine spatial resolution in the direction normal to the beam. The pulse length is adjustable. Although pulse lengths as short as 15 m are possible, it was found that such a short pulse length did not afford enough sensitivity in this case when the tornado was close enough to be within the window when it could be used, because the transmitting tube used in the radar was old and not operating at its rated full power. It was found, however, that a 30-m pulse length was long enough to achieve the sensitivity required to detect scatterers in and around the tornadoes. The volumes probed by the W-band radar were thus ∼10 m × 10 m × 30 m at the range of the tornadoes. The only near-real-time displays available in 2004 were an A-scope for determining whether or not there was sufficient sensitivity and a B-scan display for radar reflectivity, which gave us a crude idea of what the tornado looked like and how badly attenuated the signal was in the vicinity of the tornado. Properly operating the radar was thus demanding and required the operator to be very experienced.
A boresighted video camera is mounted next to the antenna, so that it is possible to correlate visual features with the radar data. In the field, the radar operator makes use of the boresighted video available in real time to set the limits of the attitude of the radar antenna before executing scans. Although the radar system was equipped with levelers, they were not used in the interest of time and also because there was a mechanical problem with them. The radar-truck platform was deployed on as level surfaces as possible, but they were not perfectly level. The amount of tilt can be estimated from the boresighted video.
The elevation angle of the radar antenna was not recorded for the dataset collected near Happy, Texas, on 5 May 2002, and had to be retrieved based on the known scan rate (Bluestein et al. 2004b). However, in the 12 May 2004 case, while the former problem did not occur, the vertically polarized signals were not recorded, owing to a bad cable/connector. It was therefore not possible to attain a maximum unambiguous Doppler velocity of ±79 m s−1 using the polarization diversity pulse pair (PDPP) technique (Pazmany et al. 1999). Consequently, the conventional pulse-pair data, which were available but folded at only ±7.8 m s−1, had to be dealiased. This tedious unfolding process was aided by using the boresighted video to estimate where the zero isodop was located. It was assumed that the zero isodop was located near the center of the tornado when the motion of the tornado was approximately normal to the line of sight of the radar beam (see Fig. 1). When the motion of the tornado had a substantial component along the line of sight, the zero isodop had to be shifted laterally. Tornado motion was determined from the location of the vortex signature associated with the tornado, from boresighted video, and from the damage tracks. (Data from the X-band radar, detailed in the following section, were also used, but only in a rough sense, since most of the scans were not exactly coincident in time with the W-band radar scans.) Even if the PDPP data had been available, it would have been used to unfold the pulse-pair data to produce the highest-quality dataset, because PDPP data are inherently noisier than conventional pulse-pair data.
Eleven high-quality RHI scans were obtained in one tornado and two RHI scans were obtained in another. Of the former, 10 were collected systematically from right to left (north to south) across the tornado. In addition, 14 other high-quality sector scans were collected in the tornadoes. Many other scans were made when the tornado was too far away, when the pulse length was too short, or when there was a problem with the scanning software. The data analyses presented in section 3 constitute the best of the dataset.
b. X-band radar (“UMass X-Pol”)
Details about this radar, also known as the UMass X-Pol, and its use are found in Pazmany et al. (2003), Junyent et al. (2004), Kramar et al. (2005), and Bluestein et al. (2007). The radar system is based on a commercial, marine radar transceiver. The antenna has a half-power beamwidth of 1.25°. The range resolution is 150 m. A staggered pulse-repetition frequency provides a maximum unambiguous velocity of ±60 m s−1 and a maximum unambiguous range of 75 km.
In 2004, two modes of data collection were available: 1) “surveillance mode,” when radar reflectivity was displayed in near-real time on a plan position indicator (PPI) display, potentially out to 120 km, and only postprocessed radar reflectivity data were available, and 2) “data collection mode,” when no data were displayed on the PPI, but postprocessed radar reflectivity, Doppler velocity, and polarization parameters were available out to a range of 30 km. The only scans collected were nearly constant elevation-angle scans around a full circle. Although levelers were not available in 2004, efforts were made, as they were with the W-band radar truck, to deploy on as level surfaces as possible. Since boresighted video was not available for the X-band truck, the amount of the truck’s tilt could not be estimated.
The radar operator switched back and forth between surveillance mode and data collection mode to verify that the target storm and its features were being probed and that the ground clutter contamination was minimized for the lowest possible elevation angle. Thus, the data record has many gaps. In addition, the computer storage was seriously limited, so that at times the operator, while in the field, had to erase old data from other cases to make room for new data. Also, owing to the way the data were processed, there were frequently gaps in some sectors when in data collection mode. The analyses of X-band radar data presented in section 3 represent some of the best and most representative of the sample of scores of scans collected on 12 May 2004.
3. Description of the data analyses
In this section the radar data and visual imagery are described for several tornadoes. How the tornadoes were related to their parent supercells is described for all of the tornadoes; detailed analyses of the structure of the tornadoes are presented for the tornado near Attica, Kansas, and the tornado southwest of Harper, Kansas (Fig. 1).
a. The tornado south of Medicine Lodge
This first tornado (whose track is not shown in Fig. 1) was spotted to the south-southeast of the deployment site of the radars, on the eastern side of Medicine Lodge, Kansas, around 1913 CDT; it is possible that the tornado had been in existence for some time, but was not viewed by the storm-intercept team. This tornado evolved into the shrinking stage and by 1924 CDT it was visible only as a narrow funnel cloud aloft and surface debris cloud. Just before the appearance of this tornado, the parent supercell had an appendage of radar echo on the rear side (Fig. 2). A persistent (at least for ∼30 min), east–west-oriented fine line was located just to the north of the parent supercell, touching or merging with it in some places. By viewing a loop of X-band radar imagery, it can be seen that the fine line moved along to the east-northeast with the supercell. It is surmised that this fine line probably marked a surface outflow boundary. The storm intercept crew noted that this supercell had formed just to the southeast of an earlier convective cell and appeared to form along the edge of its outflow (not shown; visual evidence based on cloud features). It is not known, however, if this fine line was the original boundary along which the convective storm developed.
At the time the tornado was mature (Fig. 3), the radar-echo appendage had evolved into a less pointed shape (Fig. 2c). The approximate location of the tornado relative to it was to its southeast. This tornado was relatively narrow and was near the rear edge of a cloud line that extended rearward(approximately southwestward) from its parent storm. Scatterers associated with the tornado, which was ∼6 km away, were too weak to be detected by the W-band radar. Although the appearance of the tornado, viewed from its north, was surprising to the intercept team because a well-defined wall cloud was not visible and it appeared on the far southwestern edge of the parent cloud, the X-band radar reflectivity imagery is consistent with the structure of a typical supercell in that its location was in the typical location relative to the main body of the radar echo; apparently, a well-defined hook was not visible, owing to a lack of precipitation.
b. The Attica tornado
The next in the series of tornadoes formed to the southeast of Attica, Kansas, and moved to the north or north-northwest (Fig. 1); it is henceforth referred to as “the Attica tornado.” The location of the W-band and UMass X-Pol radars was at “R1” marked in Fig. 1; the radars were at this location from ∼1950 to 2007 CDT. A wide-angle view from “X-Pol” while the tornado was forming is shown in Fig. 4, to give the reader an overall view of the tornado and the relative deployment of the radars. A clear slot was seen to the south and southwest of the developing tornado, the latter of which appeared as a rotating column of debris near the ground and a bowl-shaped funnel cloud at the cloud base. Also seen is a small funnel cloud to the north or northwest of the developing tornado. The previous tornado had appeared to dissipate, but occasionally short-lived debris whirls appeared on the ground, sometimes along with narrow condensation funnels aloft. This funnel cloud may have been a vestige of the earlier circulation. The W-band radar truck is also seen to the west, within 50 m of the UMass X-Pol.
The entire life cycle of the Attica tornado was documented with still cameras and on video (Fig. 5). It evolved in the well-known sequence (Davies-Jones et al. 2001) of a rotating debris cloud on the ground underneath a lowered cloud base with a clear slot visible to the south (Fig. 5a) and then a funnel cloud appearing aloft (Fig. 5b); the bottom of the funnel cloud propagated downward and the tornado appeared to be the widest and most mature at ∼2002–2003 CDT (Figs. 5c,d). At 2004 CDT (Fig. 5e) the tornado condensation funnel narrowed as it entered its “shrinking stage” (Fig. 5f) and eventually appeared as a narrow rope disconnected from the cloud base aloft with a wide surface debris cloud (Fig. 5g), as did another decaying tornado condensation funnel described by Bluestein et al. (1988).
The Attica tornado began at the tip of a thin band of enhanced, hook-shaped, radar echo (yellow–white, ∼17 dBZ) at the rear of the storm (Fig. 6a). This thin band, viewed in X-band imagery, looks like the “umbilical cord” observed by the W-band radar in a tornadic supercell on 15 May 1999 (Bluestein and Pazmany 2000). It has been surmised that a narrow precipitation curtain may be responsible for the thin band. The thin band was located just to the east-southeast of an appendage on the rear side of the storm, a feature that had been noted earlier (Fig. 2). The region where the Attica tornado was forming at 1952:52 CDT was marked by a narrow disc of higher (yellow, ∼20 dBZ) reflectivity, that was probably lofted dust, visible to the intercept team as rotating dust whirls (not shown).
During the beginning of the mature phase of the tornado (Fig. 5c), the X-band reflectivity imagery of the tornado displayed a coiled-up band of higher reflectivity (red, ∼45 dBZ), which was embedded within a general, circular area of moderate reflectivity (orange–brown, ∼30–40 dBZ) connected to a hook echo at the storm’s rear (Fig. 6b). A notch of relatively weak reflectivity (green, ∼10 dBZ) was located just to the northeast of the tornado-related radar echo, but just to the south of the core of the storm. All of these radar features are commonly observed in tornadoes (e.g., Wurman and Gill 2000; Bluestein and Pazmany 2000; Alexander and Wurman 2005).
The scans made by the W-band radar generally covered much narrower sectors than those covered by the X-band radar. To place the W-band radar scans in their proper context, it is useful to compare the W-band radar scans to the X-band radar when appropriate. It was possible to compare radar reflectivity from the W-band radar with that from the X-band radar, though the data were collected ∼1 min apart. The tornado at 2000:04 CDT, within several seconds of when the W-band scan was collected, looked as it did in Fig. 5b (Fig. 7a): a well-defined debris cloud near the ground was situated underneath a funnel cloud. The X-band reflectivity imagery (Fig. 7b), which is a zoomed in version of Fig. 6b, shows spiral bands of enhanced reflectivity surrounding a ring of weaker, but still enhanced ring of reflectivity, ∼600 m in diameter. This ring corresponds approximately to the width of the opaque debris cloud near the ground (Figs. 5b,c). Bluestein et al. (2007), using dual-polarization radar data, provided further evidence that the ring in this tornado was caused by debris, not by hydrometeors; it was also shown that the spiral bands were likely caused by hydrometeors. At 1959:54 CDT, ∼1 min earlier, the W-band imagery showed the spiral band as an arc of relatively high reflectivity (∼−5 dBZe; Fig. 7c), characterized by a cyclonic-vortex signature; Doppler wind speeds of ∼12 (∼−40) m s−1 on the north (south) were noted. The center of the tornado and the debris ring seen by the X-band radar were beyond the range of the W-band radar, which was restricted to be within 3.75-km range. The widely differing values dBZe may be attributed largely to the difference in the nature of the scattering, which was probably of the Rayleigh type in the case of the X-band radar, but of the Mie type in the case of the much shorter wavelength, W-band radar. Other differences could be a result of the slightly different times of the scans and calibration errors. It is believed that this comparison of the radar reflectivity seen in a spiral band of precipitation about a tornado was the first one involving both X- and W-band radars. It is important to show that both radars “see” the same features and that the locations and shapes of the features are similar, even though their radar-echo reflectivity is not.
The highlight of the data collection of the Attica tornado was a systematic set of W-band radar RHI scans through the tornado when it was mature. About 10 s before and 1 min after these RHI scans, there were sector scans at low-elevation angle through the tornado (Fig. 8). The corresponding views of the tornado when these scans were taken are seen in Fig. 9. The sector scan at 2001:33 CDT, when the tornado was mature, was taken through a well-defined debris cloud (Fig. 9a). The sector scan at 2004:15 CDT, when the tornado was in its shrinking stage (Fig. 5e), was also taken through a debris cloud, though the condensation funnel aloft was narrower than it was earlier (Fig. 9b). The sector scans passed above the level of treetops, which were estimated to be ∼10 m high and about 1/3–1/4 the distance of the tornado from the radar. At the range of the tornado, the height of the radar beam was ∼30–40 m AGL.
At 2001:33 CDT, the center of the tornado had a weak-echo hole/eye ∼200 m in diameter that was partially surrounded by a ring of reflectivity (Fig. 8a). A cyclonic couplet surrounded the eye (Fig. 8b); since the maximum approaching and receding Doppler velocities, which were separated by ∼250 m, were well outside the eye, it is unlikely that there were higher Doppler wind speeds within the eye that were not detected. It is thought that the eye is created when scatterers are centrifuged radially outward (Snow 1984; Bluestein et al. 1993; Dowell et al. 2005). It is therefore expected that systematic RHI scans across this tornado should pass through, between, and on either far side of the two peaks in Doppler wind speed. In addition, a notch of low reflectivity is seen ∼1 km beyond the tornado.
The scan at 2004:15 CDT was a full minute later than the end of the series of RHIs and was after the tornado had passed through its mature stage. It is therefore not as useful as the scan at 2001:33 CDT in characterizing the tornado. Nevertheless, it still had a 250-m-wide cyclonic-vortex signature (Fig. 8d) and its maximum Doppler wind speeds were comparable (note that the color scales used in Figs. 8b,d are different; the color scales were selected to best represent the span of velocities present in each one separately). Thus, even when entering the shrinking stage, the Doppler velocities were still high, as had been found previously in other cases (e.g., Davies-Jones et al. 1978; Bluestein et al. 1993, 2003b). The radar reflectivity depiction of the tornado during the shrinking stage (Fig. 8c) differed from that during the shrinking stage of a tornado in Nebraska on 5 June 1999 (Bluestein et al. 2003b, their Fig. 10) in that the diameter of the eye in the Attica tornado contracted (from 250 to ∼150 m), while that of the Nebraska tornado did not. The Doppler wind data collected in the Attica tornado resolved the peaks in Doppler velocity better than did the data collected in the Nebraska tornado (Bluestein et al. 2003b, their Fig. 11). In the tornado in Kansas studied by Tanamachi et al. (2007, their Figs. 6 and 17), both the radius of maximum wind (RMW), the distance from the center of the vortex at which the azimuthal wind speed is the highest, and the eye diameter decreased during the shrinking stage.
Since at 2004:15 CDT the Doppler wind speed maxima were so well resolved in the W-band dataset, the wind data were analyzed using the ground-based velocity track display technique (GBVTD) of Lee et al. (1999) in order to determine more about its structure. The GBVTD technique makes use of the assumption that the tornado is circular and can retrieve the component of the wind normal to the beam. While originally designed to analyze tropical cyclone wind data, the GBVTD technique has been applied to W-band tornado data by Bluestein et al. (2003b) and Tanamachi et al. (2007), and to X-band tornado data by Lee and Wurman (2005).
The GBVTD-estimated variation of azimuthally averaged azimuthal wind varied like a Burgers–Rott vortex (Davies-Jones et al. 2001), a version of the combined Rankine vortex in which the transition from the inner, solid-body, constant vorticity core region to the outer, potential flow region is gradual, rather than abrupt (Fig. 10a). The vorticity in the core was ∼1 s−1, which is characteristic of tornadoes. The RMW was ∼110 m, which is consistent with the ∼250-m width of the Doppler wind speed maxima couplet (Fig. 8d). The maximum wind was estimated as ∼50 m s−1, which is consistent with the damage estimate of F2 intensity (Fig. 1), even though the radar-estimated wind speeds were ∼30–40 m AGL. The circulation of the tornado increased with radius, but leveled off to ∼40 000–53 000 m2 s−1 beyond the RMW. In comparison, Dowell and Bluestein (2002, see their Fig. 11) found in airborne Doppler radar analyses of a tornado in the Texas Panhandle in 1995, that in most, but not all instances a circulation associated with the tornado leveled off to ∼150 000–350 000 m2 s−1, beyond ∼1.5-km radius; in their analyses, however, the circulation represented a scale of motion longer than that of the tornado itself. However, down to radii of 400–500 m, which is just a few hundred meters beyond the radius of measurements in the tornado shown in Fig. 10, the circulation was also ∼50 000 m2 s−1.
The GBVTD-estimated radial wind component was inward within ∼50 m of the center, and outward beyond (Fig. 10a). Thus, there was low-level divergence (divergence was computed as ∂υr/∂r + υr/r, where υr is the radial wind component and r is the distance from the center of the vortex) within the RMW; if this were indeed representative of the wind field near the ground, then it can be inferred kinematically that there was sinking motion within the RMW. While this finding is consistent with the shrinking and decaying stage of a tornado, it must be viewed with caution because the retrieved radial inflow within 50 m has some uncertainty, owing to the weaker reflectivity near the eye (Figs. 10b and 8c) and to the possible centrifuging of larger debris particles out at greater distance from the center of the tornado (Dowell et al. 2005), which would contribute to radial outflow there. As in dust devil vortices (Bluestein et al. 2004a), the maximum radar reflectivity was found within the RMW. It has been speculated that debris could have been transported radially inward in the friction layer below and then lofted. The divergence might also represent the top of the frictionally induced secondary circulation and therefore not necessarily indicative of sinking motion, but instead of decreasing rising motion with height.
Beginning at 2001:44 CDT, 10 RHIs were taken through the tornado, ending with the last one beginning at 2003:27 CDT; each scan took ∼10 s. Since the tornado was moving nearly normal to the line of sight (and the radar beam), the scans were tilted slightly (Fig. 11); upward (downward) scans produced quasi-vertical cross sections that were tilted to the left, south (right, north) with height. The tilt of the scans is in sharp contrast to the tilt of the scans of the Happy, Texas, tornado on 5 May 2002, when the motion of the tornado was predominantly along the line of sight (and the radar beam). In the Happy case, the vertical tilt of the quasi-vertical cross sections was due mainly to along-beam motion. It is very difficult, owing to tornado motion, to capture untilted cross sections; unless the tornado were nearly stationary or the scans very rapid, it would be impossible to obtain perfectly vertical cross sections. Another source of tilt is the tilt of the radar-truck platform itself. It is seen from the boresighted video (Figs. 7a, 9 and 12) that the radar platform was tilted in the north–south direction (approximately normal to the direction the truck was facing) slightly, ∼2.5°–3° (upward to the south–downward to the north); owing to the tilt of the truck platform, the downward scans were more vertically oriented than the upward scans. It is not known how much, if any, the radar platform was tilted in the east–west direction (along which the truck was facing). In addition, it is also possible that the tornado itself is tilted.
It was determined through photographic analysis of still photographs and captured boresighted video frames that the bottom of the RHI scans was on the average ∼8 m AGL. The uncertainty in this estimate (∼±5 m) is due in large part to errors in our estimates of the heights of trees and their range; the actual bottom of the scan could have been as much as 5 m higher or lower.
Each cross section encompassed ∼100 m in the horizontal (∼10 m s−1 tornado motion × 10 s per scan), so that each one can be thought of as representing a vertical slab ∼100 m × 1100 m (the approximate top of each scan). The northernmost (rightmost) scan passed just outside the edge of the semitranslucent part of the debris cloud and well to the right of the condensation funnel; the southernmost (leftmost) scan passed just inside the semitranslucent debris cloud. It was thus assured that most of the tornado’s core and a region outside the core were sampled. The appearance of the tornado, which was mature, did not change much over the duration of the RHIs (Fig. 12); it will therefore be assumed that the tornado was approximately steady during the sequence of vertical scans.
Three major features are prominent in the reflectivity imagery of the vertical cross sections through the Attica tornado (Fig. 13). First and foremost, a weak-echo hole/eye, which was approximately 600 m in diameter at heights above ∼400 m AGL, is evident in scans 3–8, which extends outward from the condensation funnel by as much as 100–200 m (Fig. 11). The eye, however, does not extend much, if any, beyond the edge of the outer, semitranslucent debris cloud. The eye extends down the farthest to the ground in scan 5, which is the scan closest to the center of the tornado; this finding is consistent with geometrical considerations, if the eye is in fact coincident with the center of the tornado vortex. Weak-echo holes are commonly observed in tornadoes (e.g., Wakimoto and Martner 1992; Wurman and Gill 2000; Bluestein and Pazmany 2000)
The second major feature in the vertical cross sections of reflectivity is a quasi-horizontal, radial bulge in the eye that appears in scans 6–8, and by itself in scans 9–10 (Fig. 13). This bulge, which is ∼900 m long, descends with increasing scan number [i.e., with increasing distance to the south (left)]. At ∼600 m AGL in scan 6, just to the south (left) of the center of the tornado, it falls off to ∼250 m AGL in scan 10, far (∼400 m) to the south (left) of the center of the tornado. The height of the bulge nearest to the center of the tornado coincides approximately with the top of the visible debris cloud (Fig. 5c). It is believed that this feature is unique to this dataset; it was not apparent in the Happy tornado (Bluestein et al. 2004b). Although it is not clear what caused the bulge, a possible explanation is offered in the summary and discussion section.
The third major feature evident in the vertical cross sections is the notch of low radar reflectivity seen in scans 1–5, and possibly very near the ground in scans 6–7 (Fig. 13). The notch is thus most prominent to the north of the tornado. In scans 1–4, it is bounded to the west by a relatively smooth wall of higher reflectivity that curves upward with height to the east (to the right in Fig. 13), in the direction of the tornado. The inner wall of higher reflectivity is more erect and more ragged. The spacing between the walls decreases with height and closes up in scan 1 the highest, at ∼800 m AGL. The center of the notch near the ground is ∼1.1 km from the center of the tornado (e.g., see Fig. 13, scan 3). The X-band radar reflectivity image from ∼1 min earlier (Fig. 6b) has a weak-echo notch on the northeast, north, and northwest sides of the tornado. In Fig. 7b, the notch is evident ∼1.5 km away from the center. If the air in the weak-echo notch seen in Figs. 6b and 7b had been advected cyclonically about the tornado and inward radially slightly, then ∼1 min later, when scans 1–4 were taken, then the notches seen in Figs. 6b, 7b, and in Fig. 13 are the same feature. Weak-echo notches are also commonly observed in tornadoes (e.g., Wurman and Gill 2000; Bluestein and Pazmany 2000), though they have not been analyzed in such detail. It is beyond the scope of this paper to explain the three-dimensional shape of the weak-echo notch, which is described here with what is probably, to date, the highest spatial resolution. To explain the tilt of the notch’s walls and the vertical cross section of the notch would require precise knowledge of the three-dimensional wind field and precipitation field as a function of time. It is speculated that the three-dimensional structure of the notch reveals aspects of the low-level wind that might be of importance in describing tornado structure.
The vertical cross sections of Doppler velocity across the tornado are qualitatively consistent with a vortex centered in the eye (Fig. 14). The highest receding Doppler velocities are located to the north (right) of the center of the tornado at low levels in scans 2–3 (Fig. 15a), which were to the north (right) of the condensation funnel below ∼500 m AGL; the highest approaching Doppler velocities are located to the south (left) of the center of the tornado at low levels in scans 6–7 (Fig. 15a), which were to the south (left) of the condensation funnel below ∼500 m AGL. If the core radius of the tornado near the ground did not change (in Fig. 10 it is seen that the core radius was ∼110 m) between ∼2002 and 2004 CDT, then the core of the tornado near the ground was sampled best by scan 6; the core was best sampled well above the ground by scans 3–4 (Fig. 11).
Since the RHIs through the tornado were tilted relative to the ground (Fig. 11), some of the apparent vertical variations in Doppler velocity and speed (Fig. 15) represent radial variations in wind. For example, if a scan cut radially inward inside (outside) the core radius of the vortex, then one would expect there to be some decrease (increase) in wind speed superimposed on any existent vertical variation. Since alternate scans cut in the opposite radial direction (Fig. 11), one would expect to find that if there were significant contamination in vertical variations from radial variations, that it would be manifest as different shapes in the vertical profiles of Doppler velocity. In Fig. 11 it is seen that the downward scans (i.e., those that are evenly numbered) were tilted less than the upward scans, so that the former should be less contaminated than the latter.
From Fig. 11 it is thus seen that scans 4 and 6 were probably the most representative vertical scans because they passed through the core, inside it, and were downward scans. The way in which the azimuthal wind speeds varied with height in the Attica tornado are therefore best determined by considering the vertical variation of the Doppler wind speed shown in Fig. 15b for scans 4 and 6. It is seen that for scan 6, the Doppler wind speed increased with height rapidly from ∼60 m s−1 near the ground (which is consistent with F2 damage), reaching a maximum of ∼77 m s−1 between ∼25 and 75 m AGL. For scan 4, the wind speed increased from ∼10 m s−1 near the ground to ∼20 m s−1, ∼50 m AGL. Both scans 4 and 6 were apparently inside the eye above ∼100–150 m AGL, where no wind data were available. Since the wind speeds from scan 4 were relatively low, it is likely that the bottom of scan 4 passed through the core, very close to its center; since the wind speeds from scan 6 were relatively high, in fact even higher than those seen in Fig. 10, which is valid for a time a few minutes later and at a lower altitude, it is likely that scan 6 best represents the vertical variation in the azimuthal wind with height near the core of the tornado on the approaching velocity side. The W-band radar probably did indeed sample the vertical variation in wind in the surface friction layer in scan 6. If the scan were just inside the core (Fig. 11), then a downward scan would have moved toward larger radii, where the wind speed would have increased (Fig. 10), rather than decreased, as was observed. It therefore appears that the rate of increase of wind speed with height in the surface friction layer may have been even greater than that indicated in Fig. 15. However, any conclusions about the rate of change of wind speed with height near the ground must be viewed with caution since the wind velocity gradients were large near the ground, and consequently there was the greatest potential there for unfolding errors.
From Fig. 11, it is seen that scans 2 and 8 were probably the next most representative vertical scans because they passed outside the core and were downward scans. The Doppler wind speed in scan 2 (on the receding velocity side of the tornado) showed an increase from ∼14 m s−1 near the ground to ∼40 m s−1 just 10–20 m AGL (Fig. 15b). The Doppler wind speed in scan 8 (on the approaching velocity side) also showed an increase with height, but less than that in scan 2: the wind speed increased from ∼53 m s−1 near the ground to ∼62 m s−1 and ∼20 m AGL. Again, the W-band radar probably did in fact sample the vertical variation in wind in the surface friction layer in scans 2 and 8, outside the tornado core; the same cautionary note is sounded, owing to potential unfolding errors where the wind velocity gradient is large.
To further highlight the structure of the tornado, the Doppler velocity field displayed for scan 5 is reproduced in expanded format in Fig. 16. Data were available down to as low as a depth of one or two pixels at the center, which was within ∼10 m or so of the ground. Two vortex signatures in the vertical plane (i.e., representing horizontal vorticity) are evident on the far side of the tornado (yellow–white over blue–purple). One is centered ∼200 m AGL and the other is centered ∼400 m AGL. Each indicates radial inflow of ∼21 m s−1 underneath radial outflow of ∼0–7 m s−1. These couplets are spaced ∼250 m apart in the vertical. They did not appear to be symmetrical about the tornado, though there was a quasi–wavelike vertical variation in Doppler velocity on the near side, where inward flowing scatterers of ∼27 m s−1 decreased to ∼15 m s−1, quasi-periodically, every ∼100 m. While it is possible that unfolding errors could be responsible for this quasi-periodicity, it is unlikely because the amplitude of the fluctuations (∼12 m s−1) was not the same as the Nyquist interval. The vortex signatures might represent horizontal vortices along the edge of the tornado, as, for example, in the large eddy simulations of Lewellen et al. (2000; their Fig. 5). The lowest Doppler velocity minima decreased in height with radius and extended far beyond the core of the tornado (out to ∼800–900 m from the center, at least). It is not known whether this “jet” of outflow represents the vertical variation of the wind in the planetary boundary layer east of the tornado, for wind veering with height in the near-storm environment in the lowest several hundred meters, or whether it is related to the tornado. The soundings at 1900 at Lamont, Oklahoma (Fig. A2), about 100 km to the southeast, and at Norman, Oklahoma (not shown), far to the south-southeast of the tornado, exhibited a gradual veering of the wind southeasterly–southwesterly ∼1.5–2 km AGL; the corresponding sounding at Topeka, Kansas, far to the northeast of the tornado, exhibited sharp veering of the wind in the lowest few hundred meters southeasterly–southwesterly (not shown). A similar jetlike feature was noted in the Happy tornado on 5 May 2002 (Bluestein et al. 2004b, their Fig. 10).
c. The tornado subsequent to the Attica tornado, southwest of Harper
After the Attica tornado had dissipated, the radar trucks moved to the east to keep up with the supercell that had produced the Attica tornado and its predecessor. Another tornado formed (Fig. 17), which was probed by both the W-band radar and UMass X-Pol when it was southwest of Harper. Although the dataset collected was not as comprehensive as the former, it is useful to discuss it briefly, to look for similarities and differences with the former. The locations of the radar trucks were “R2” and “R3,” both located to the northeast of the tornado (Fig. 1). This tornado was mostly over an open field and was rated only as F0 intensity, despite its appearance with a well-defined condensation funnel (Fig. 17a) and debris cloud (Figs. 17a,b).
This tornado was associated with a tiny eye at the tip of a hook echo and a weak-echo notch to its west, northwest, and north (Fig. 18a) at 2021:28 CDT and to its north and northeast at 2038:24 CDT (Fig. 18b); a weak-echo notch and eye were also seen in reflectivity data of the Attica tornado. Even more tornadoes formed after this one (Fig. 1), but they will not be discussed because the UMass radars did not collect any high-quality data in them. The only unique reflectivity image seen showed that the hook echo and eye/hole had evolved such that an anticyclonic hook echo was located inside the eye/hole (Fig. 18c).
The W-band radar imagery at 2012:25 and 2024:19 CDT, which corresponds in time approximately to the photographs of the tornado seen in Fig. 17, are considerably different. The former exhibits a 400-m-wide eye and suggestions of a relatively weak cyclonic-vortex signature about the eye (Figs. 19a,b). Maximum Doppler velocities are only ∼15 m s−1 away from the radar, though data from the region of maximum inbound velocities were not available and might have contained higher wind speeds. The latter shows a much narrower, 100-m-wide eye, with a cyclonic-vortex signature having maximum Doppler velocities of ∼40 m s−1 in the receding direction (Figs. 19c,d). A GBVTD analysis of the latter indicates that the RMW was only 70 m and the maximum azimuthal wind speed was ∼37 m s−1 (Fig. 20a), which is consistent with F1 intensity, not F0 intensity as estimated by the NWS (Fig. 1). It is likely that the lack of any damaged structures to inspect led to the underestimate. It is also possible that the winds at the ground were weaker than those sampled by the radar beam, which was elevated above the ground.
In this tornado, the radial profile of azimuthal wind at 2024:19 CDT was like that of a Burgers–Rott vortex, as was that of the Attica tornado. Owing to the weaker winds, and smaller RMW, the circulation was also weaker (Fig. 20b). The radial wind increased with radius up to around the RMW and then decreased with radius. Associated with this radial wind profile, there was a peak in divergence inside the RMW and little divergence around and just beyond the RMW. It may therefore be inferred that, at this time, sinking motion characterized the tornado, which shortly thereafter dissipated. If the divergence estimates were contaminated significantly as a result of the centrifuging of larger pieces of debris, then divergence would have increased with radius at least until the RMW, which is not what was found.
One vertical scan was collected just after this time at 2024:43 CDT (Fig. 21), which passed through the center of debris cloud and condensation funnel (not evident in Fig. 17b). Near the ground, the scan passed just to the right (northwest) of the center, while aloft the scan passed just to the left (southeast) of the center. Only the lower part of the 200–250-m-wide eye is evident (Fig. 22), which is much wider than the 100-m-wide eye seen in Fig. 19c; the width of the eye may have widened between 2024:19 and 2024:43 CDT. It is also possible that if the scan depicted in Fig. 19c was at the lowest scan level seen in Fig. 22a, then the width of the eye was only ∼150 m. Doppler velocities near the ground were not available, owing to the weak-echo hole/eye. Just above the eye, the radar beam sliced through the left (southeastern) side of the tornado vortex, where approaching velocities ∼−40 m s−1 (purple) are evident.
The most noteworthy aspect of the vertical profile of Doppler velocities is the appearance of two Doppler signatures for horizontal vortices on the far side of the tornado (yellow–white over green), similar to the ones seen in the Attica tornado (Fig. 16). The signatures at 2024:43 CDT were located at ∼75 m and at 200 m AGL, which are lower levels than those at 2002:26 CDT.
4. Summary and discussion
For the first time, tornado data were collected nearly simultaneously from the same location with both W- and X-band Doppler radars; the interpretation of the data was aided with boresighted video and still photographs. The main findings based on analyses of the dataset are as follows: 1) The Attica tornado had a weak-echo hole/eye that closed up within approximately 10 m or so of the ground. Such a structure is similar to some simulated by Dowell et al. (2005). The maximum radar reflectivity was found inside the RMW. These findings are consistent with the inward advection of small scatterers in the surface friction layer (Figs. 23a,b). The eye extended outward from the edge of the condensation funnel by as much 1–2 times the core radius; however, the eye did not extend significantly beyond the edge of the outer, translucent debris clouds. A bulge in the eye descended with height with increasing distance from the center of the tornado on its south side. Evidence was found of horizontal vortices at low levels along an edge of two of the tornadoes. Similar features have been seen in large eddy simulations of tornado-like vortices (Fig. 23c). The bulges in the eye in the Attica tornado, however, were seen on the south side of the tornado, while the horizontal vortex signatures were found on the north side. The bulges can therefore not be related to any azimuthally symmetric vertical circulations below. It is suggested that the bulge of weak reflectivity may have represented scatterer-free air forced radially outward from the tornado center by a circulation at the level of the bulge. However, this explanation is highly tentative because we cannot explain why the bulge would appear only on one side of the tornado. Furthermore, we cannot verify this hypothesis because Doppler wind data were not available owing to very weak radar reflectivity (lack of scatterers). 2) Scans penetrated into the surface friction layer both inside the core and outside the core. In the most representative scan inside the core, it was found that the Doppler velocity increased by almost 30% (from 60 to 77 m s−1) in the lowest 25 m AGL. Outside the core, the Doppler velocity increased by 180% (from 14 to 40 m s−1) in the lowest 10–20 m. Sharp gradients in wind velocity are also found in high-resolution numerical simulations (e.g., Lewellen et al. 2000) and in other Doppler radar observations (e.g., Alexander and Wurman 2005). 3) The radial profiles of azimuthal winds were similar to that of a Burgers–Rott vortex, as has been found in other tornadoes (e.g., Wurman and Gill 2000; Bluestein et al. 2003b; Lee and Wurman 2005; Wurman and Alexander 2005; Tanamachi et al. 2007). 4) Even in the shrinking stage, the winds in the tornado were still relatively strong. The diameter of the weak-echo hole/eye decreased during the shrinking stage. 5) Both the W- and X-band radars detected a weak-echo notch that was wrapping around the tornado. The notch was widest at the ground and closed up aloft. The far side of the notch sloped radially inward with height; the near side of the notch was erect. It is not known whether the notch represented inflow air or precipitation-void air originating in the rear-flank downdraft. 6) Both radars detected a spiral band of precipitation surrounding the tornado (e.g., Wurman and Gill 2000; Bluestein and Pazmany 2000); the radar reflectivity from the W-band radar was much weaker than that from the X-band radar, because the nature of the scattering was different.
It is hoped that more tornado datasets will be collected in which simultaneous W- and X-band radar data are available, along with visual documentation. These datasets will serve to corroborate the findings from the 12 May 2004 case and also to document other structures associated with tornadoes.
Acknowledgments
This work was supported by NSF grant ATM-0241037 to OU and ATM-0000592 to the University of Massachusetts. The second and third author operated the W-band and X-band radars, respectively, in the field. Dan Dawson and Brad Barrett participated in field operations as drivers and navigators. The authors thank Mark Laufersweiler (OU) for his computer assistance. Wen-Chau Lee and Mike Bell (NCAR/ATD) provided software and assistance for the GBVTD processing. Jeff Snyder (OU) edited one of the RHI scans, but it was not reproduced in this manuscript. Three anonymous reviewers provided very useful comments. Part of this work was done while the first author was a visiting scientist in the Mesoscale and Microscale Meteorology Division (MMM) at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado. NCAR is supported by the National Science Foundation.
REFERENCES
Alexander, C. R., and J. Wurman, 2005: The 30 May 1998 Spencer, South Dakota, storm. Part I: The structural evolution and environment of the tornadoes. Mon. Wea. Rev., 133 , 72–96.
Bluestein, H. B., 1999: A history of storm-intercept field programs. Wea. Forecasting, 14 , 558–577.
Bluestein, H. B., and A. L. Pazmany, 2000: Observations of tornadoes and other convective phenomena with a mobile, 3-mm wavelength, Doppler radar: The spring 1999 field experiment. Bull. Amer. Meteor. Soc., 81 , 2939–2951.
Bluestein, H. B., E. W. McCaul Jr., G. Byrd, and G. Woodall, 1988: The unususal dissipation of a tornado funnel. Mon. Wea. Rev., 116 , 950–952.
Bluestein, H. B., W. P. Unruh, J. LaDue, H. Stein, and D. Speheger, 1993: Doppler radar wind spectra of supercell tornadoes. Mon. Wea. Rev., 121 , 2200–2221.
Bluestein, H. B., W. P. Unruh, and J. LaDue, 2003a: Mobile Doppler radar observations of a tornado in a supercell near Bassett, Nebraska, on 5 June 1999. Part I: Tornadogenesis. Mon. Wea. Rev., 131 , 2954–2967.
Bluestein, H. B., W-C. Lee, M. Bell, C. C. Weiss, and A. L. Pazmany, 2003b: Mobile Doppler radar observations of a tornado in a supercell near Bassett, Nebraska, on 5 June 1999. Part II: Tornado-vortex structure. Mon. Wea. Rev., 131 , 2968–2984.
Bluestein, H. B., C. C. Weiss, and A. L. Pazmany, 2004a: Doppler radar observations of dust devils in Texas. Mon. Wea. Rev., 132 , 209–224.
Bluestein, H. B., C. C. Weiss, and A. L. Pazmany, 2004b: The vertical structure of a tornado near Happy, Texas, on 5 May 2002: High-resolution, mobile, W-band Doppler radar observations. Mon. Wea. Rev., 132 , 2325–2337.
Bluestein, H. B., M. M. French, S. Frasier, K. Hardwick, F. Junyent, and A. L. Pazmany, 2007: Close-range observations of tornadoes in supercells made with a dual-polarization, X-band, mobile Doppler radar. Mon. Wea. Rev., in press.
Davies-Jones, R. P., D. W. Burgess, L. R. Lemon, and D. Purcell, 1978: Interpretation of surface marks and debris patterns from the 24 May 1973 Union City, Oklahoma tornado. Mon. Wea. Rev., 106 , 12–21.
Davies-Jones, R. P., R. J. Trapp, and H. B. Bluestein, 2001: Tornadoes. Severe Convective Storms, Meteor. Mongr., No. 50, Amer. Meteor. Soc., 167–221.
Dowell, D. C., and H. B. Bluestein, 2002: The 8 June 1995 McLean, Texas, storm. Part II: Cyclic tornado formation, maintenance, and dissipation. Mon. Wea. Rev., 130 , 2649–2670.
Dowell, D. C., C. R. Alexander, J. M. Wurman, and L. J. Wicker, 2005: Centrifuging of hydrometeors and debris in tornadoes: Radar-reflectivity patterns and wind-measurement errors. Mon. Wea. Rev., 133 , 1501–1524.
Junyent, F., A. L. Pazmany, H. B. Bluestein, M. R. Kramar, M. M. French, C. C. Weiss, and S. Frasier, 2004: Dual-polarization, X-band, mobile Doppler radar observations of hook echoes in supercells. Preprints, 22d Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., CD-ROM, P11.7.
Kosiba, K. A., R. J. Trapp, and J. M. Wurman, 2005: The 12 May 2004 Harper, KS tornado: Analysis of DOW radar observations of the low level wind field. Preprints, 32d Conf. on Radar Meteorology, Albuquerque, NM, Amer. Meteor. Soc., CD-ROM, P15R.11.
Kramar, M. R., H. B. Bluestein, A. L. Pazmany, and J. D. Tuttle, 2005: The “Owl Horn” radar signature in developing Southern Plains supercells. Mon. Wea. Rev., 133 , 2608–2634.
Lee, W-C., and J. Wurman, 2005: Diagnosed three-dimensional axisymmetric structure of the Mulhall tornado on 3 May 1999. J. Atmos. Sci., 62 , 2373–2393.
Lee, W-C., B. J-D. Jou, P. L. Chang, and S. M. Deng, 1999: Tropical cyclone kinematic structure retrieved from single-Doppler radar observations. Part I: Doppler velocity patterns and the GBVTD technique. Mon. Wea. Rev., 127 , 2419–2439.
Lewellen, D. C., W. S. Lewellen, and J. Xia, 2000: The influence of a local swirl ratio on tornado intensification near the surface. J. Atmos. Sci., 57 , 527–544.
Oye, R., C. K. Mueller, and S. Smith, 1995: Software for radar translation, visualization, editing, and interpolation. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Soc., 359–361.
Pazmany, A. L., J. Galloway, J. B. Mead, I. Popstefanija, R. E. McIntosh, and H. B. Bluestein, 1999: Polarization diversity pulse pair technique for millimeter-wave Doppler radar measurements of severe-storm features. J. Atmos. Oceanic Technol., 16 , 1900–1911.
Pazmany, A. L., F. Junyent, H. B. Bluestein, and M. Kramar, 2003: Quantitative rain measurements with a mobile, X-band, polarimetric Doppler radar. Preprints, 31st Conf. on Radar Meteorology, Seattle, WA, Amer. Meteor. Soc., 858–859.
Snow, J. T., 1984: On the formation of particle sheaths in columnar vortices. J. Atmos. Sci., 41 , 2477–2491.
Tanamachi, R. L., H. B. Bluestein, S. S. Moore, and R. P. Madding, 2006: Infrared thermal imagery of cloud base in tornadic supercells. J. Atmos. Oceanic Technol., 23 , 1445–1461.
Tanamachi, R. L., H. B. Bluestein, W-C. Lee, M. Bell, and A. Pazmany, 2007: Ground-based velocity track display (GBVTD) analysis of W-band Doppler radar data in a tornado near Stockton, Kansas, on 15 May 1999. Mon. Wea. Rev., in press.
Wakimoto, R. M., and B. E. Martner, 1992: Observations of a Colorado tornado. Part II: Combined photogrammetric and Doppler radar analysis. Mon. Wea. Rev., 120 , 522–543.
Weisman, M. L., and J. B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy. Mon. Wea. Rev., 110 , 504–520.
Wurman, J., 2002: The multiple vortex structure of a tornado. Wea. Forecasting, 17 , 473–505.
Wurman, J., and S. Gill, 2000: Finescale radar observations of the Dimmit, Texas (2 June 1995) tornado. Mon. Wea. Rev., 128 , 2135–2164.
Wurman, J., and C. R. Alexander, 2005: The 30 May 1998 Spencer, South Dakota, storm. Part II: Comparison of observed damage and radar-derived winds in the tornadoes. Mon. Wea. Rev., 133 , 97–119.
Wurman, J., J. M. Straka, and E. N. Rasmussen, 1996: Fine-scale Doppler radar observations of tornadoes. Science, 272 , 1774–1777.
Wurman, J., J. M. Straka, E. N. Rasmussen, M. Randall, and A. Zahrai, 1997: Design and deployment of a portable, pencil-beam, pulsed, 3-cm Doppler radar. J. Atmos. Oceanic Technol., 14 , 1502–1512.
APPENDIX
Storm Environment
This appendix is included to provide the reader with a broad synoptic overview of the convective storm that spawned the tornadoes discussed in the main section of this paper. The convective storm that spawned the tornadoes near Attica and Harper, Kansas, developed from multicell convection (radar images not shown) that had been initiated ∼1730 CDT west of Medicine Lodge, Kansas. At 1800 CDT, a cold front/outflow boundary that was situated across Kansas lay just to the north of Medicine Lodge (Fig. A1a) and a dryline that had progressed eastward into far northwestern Oklahoma intersected the boundary somewhere to the west of Medicine Lodge. East of the intersection point, the cold front/outflow boundary was characterized by a wind shift from southeasterly on the warm side to northeasterly on the cold side. West of the intersection, the boundary was characterized by a wind shift from southerly–southwesterly on the warm, dry side behind the dryline, to southeasterly on the cooler, moister side; however, the temperatures just north of this boundary had warmed up as a result of insolation so that most of the thermal gradient lay to the north and northeast of this boundary, and the winds had shifted into the southeasterly direction.
If the 1900 CDT (0000 UTC) special sounding at Lamont, Oklahoma (Fig. A2), which is about 100 km to the southeast and upstream with respect to the surface wind of the tornadic storm, and released less than 1 h before 1900 CDT, is representative of the storm environment, then convective temperature was attained at Medicine Lodge and a parcel from the boundary layer would have had CAPE ∼3700 J kg−1. Furthermore there was virtually no capping inversion to inhibit the formation of convective clouds. For a 6-km wind of ∼17.5 m s−1 from the west-southwest and a southeasterly wind at the surface of ∼5 m s−1, as indicated by the sounding and the winds plotted on the surface map, there was ample vertical wind shear (at least 20 m s−1 over the lowest 6 km) and potential instability for the development of supercells (Weisman and Klemp 1982). The vertical shear was even stronger in the boundary layer and concentrated in the lowest 2 km, since the winds increased in speed and veered rapidly over the lowest 2 km; there was little vertical shear between 2 and 6 km AGL, where the winds were from about the same west-southwesterly direction and increased with height at most only 2.5 m s−1. It is further likely that in the vicinity of the boundary and especially north of it, the surface winds were backed even more and the vertical shear even stronger. At 1900 CDT (0000 UTC) the surface wind at Medicine Lodge, just to the west of the tornadic supercell, had shifted to the north behind the front/outflow boundary, while upstream at Enid, Oklahoma, the surface wind had backed more to the southeasterly direction (Fig. A1b). Thus, by 1900 CDT, the vertical shear in the boundary layer may have increased even more.
Synoptic-scale forcing in the midtroposphere was very weak, owing to a nearly uniform and straight southwesterly current in a region of weak temperature gradients (Fig. A3), and therefore played little direct role in storm formation. Since there was a well-defined trough at 500 hPa over Idaho, Utah, and the Arizona–New Mexico border, the location of strong forcing associated with the wave considerably lagged in the south-central Kansas region.
In summary, the environment of the Attica storm was supportive of supercells, very strong vertical shear was concentrated in the boundary layer, very little if any vertical lift was necessary to initiate convection, and it is likely that a surface front/outflow boundary played roles in determining where the storm was initiated and its subsequent character. It is beyond the scope of this paper to link the precise details of storm initiation and evolution and tornado structure and evolution to the environmental parameters.
Tornado tracks (solid lines) and estimated Fujita scale rating of tornadoes on 12 May 2004, as determined by the National Weather Service, Wichita, KS. Also shown are three of the four main deployment sites of the UMass mobile radars (R1, R2, and R3) along the main highways.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Radar reflectivity imagery on the storm scale from the UMass, mobile, X-band, Doppler radar (UMass X-Pol) at (a) 1853:17, (b) 1855:24, and (c) 1917:02 CDT 12 May 2004 at a low-elevation angle. In (c), the small circle marks the approximate position of the first tornado scanned. The radar was located 1.9 miles (3 km) east of the junction of Highways 281 and 160, on 160, on the east side of Medicine Lodge, KS (not shown in Fig. 1). Range markers are shown in white every 2 km in (a) and (b), and every 1 km in (c). Color scales at the left-hand side of each panel indicate the radar reflectivity in dBZ. Features indicated on each panel include a fine line, and appendage, and swaths of beam blockage. The arrow in (c) points to a feature that is probably the successor to the appendages seen in (a) and (b).
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
The first tornado scanned, to the south-southeast of the position noted in the caption for Fig. 2, atapproximately 1920 CDT 12 May 2004. The inverted “V” shape is a bird. (Photograph courtesy of H. Bluestein)
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
The beginning of the second tornado (the “Attica tornado”) scanned, south of Attica, KS, as seen from a vantage point approximately 3.5 miles (5.6 km; see location “R1” in Fig. 1) east of Attica, on Highway 160, at approximately 1958 CDT 12 May 2004. This wide-angle view shows the beginning of the tornado as a rotating column of dust underneath a cone-shaped lowered cloud base, a clear slot to its (left) south, a funnel cloud, and the W-band radar truck. The photographer seen in the lower right is using a thermal, infrared camera (Tanamachi et al. 2006). (Photograph courtesy of H. Bluestein)
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Images depicting the evolution of the Attica tornado during its entire lifetime at approximately (a) 1959, (b) 2000, (c) 2002, (d) 2003, (e) 2004, (f) 2006, and (g) 2006 CDT 12 May 2004. The view is to the west from the same location as the photograph shown in Fig. 4. In (f) and (g), the W-band radar truck is seen in the lower left; in (f), the X-band radar antenna is also seen, in the upper right. In (c), the approximate dimensions of the tornado condensation funnel near cloud-base level, the width of the dense debris cloud at the ground, and the depth of the dense debris cloud, based on photogrammetric analysis, are indicated. (Photographs (a)–(e) and frames captured from a hand-held video camera (f) and (g) are courtesy of H. Bluestein)
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Radar reflectivity imagery at a low-elevation angle (exact elevation angle not known) on the storm scale, from UMass X-Pol, at (a) 1952:52 and (b) 2001:09 CDT 12 May 2004, from the same vantage point as in Fig. 4 (“R1” in Fig. 1). The image in (a) is representative of the period when the Attica tornado was developing; the image in (b) is representative of the time when the Attica tornado was nearing maturity (see Fig. 5). Color scales at the left-hand side of each panel are the radar reflectivity in dBZ. Range markers are shown in white at (a) 2- and (b) 1-km intervals; the ranges plotted in both (a) and (b) are in km. Features highlighted include the location of the developing Attica tornado (small yellow disc), thin band, and swaths of (a) beam blockage and (b) weak-echo notch.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
A comparison of the visual view of the Attica tornado with images from data collected by the X- and W-band radars, when it was nearing maturity. Although the times for the imagery shown are not the same, they are within ∼1 min of each other and represent the best that could be obtained for comparison purposes. (a) Captured frame from the boresighted video camera on the W-band radar at 2000:04 CDT (see Fig. 5b for a clearer, wider-scale view from a hand-held camera at approximately the same time); (b) radar-reflectivity image from UMass X-Pol at 2001:09 CDT at a low-elevation angle (exact elevation angle not known); range markers are shown in white every 200 m; selected range markers are also indicated (km) in black every 400 m, to aid the viewer; color scale for reflectivity is shown in dBZ at the right-hand side; the black circle marks the inner debris ring; the small white circle marks the center of the tornado vortex; (c) radar reflectivity (Z) from the W-band radar at 1959:54 CDT; range markers are shown in white every 200 m and indicated in km; the small white circle marks the location of the center of a circle defined by the arc of relatively high reflectivity at ∼3.2–3.6 km; this arc corresponds to a spiral band of high reflectivity in (b); color scale for reflectivity is shown in dBZe; (d) corresponding Doppler velocity (V) to the reflectivity imagery in (c); color scale for Doppler velocity is shown in m s−1 at bottom.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Quasi-horizontal representations of the Attica tornado (sector scans) (a), (b) just before and (c), (d) after the detailed vertical cross-section scans shown in Figs. 13, 14 and 16. Imagery from data collected by the W-band radar. (a) Radar reflectivity (Z) and (b) Doppler velocity (V) at 2001:33 CDT; (c), (d) as in (a), (b), respectively, but at 2004:15 CDT. Range markers are shown in white every 100 m and indicated in km. Features highlighted include (a) a weak-echo notch and (a), (c) weak-echo hole/eye. The white circles in (b), (d) enclose the Doppler cyclonic vortex signature–couplet associated with the tornado. Color scales for reflectivity (Doppler velocity) shown at the bottom in dBZe (m s−1).
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Captured frames from the boresighted video camera on the W-band radar of the Attica tornado at (a) 2001:36 and (b) 2004:27 CDT, which correspond to the radar data shown in Fig. 8.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Structure and properties of the Attica tornado at 2004:15 CDT as determined from W-band radar data at the lowest elevation angle possible (cf. Figs. 8d and 9b). (a), (b) All of the variables plotted as function of radius (distance from the center of the vortex) of the tornado. (a) Azimuthal wind component (m s−1) estimated by a GBVTD analysis (solid line connected by dots) and of a Burgers–Rott vortex fit to the profile of GBVTD-determined azimuthal winds (solid line connected by triangles); radial wind component estimated by the GBVTD analysis (m s−1; dashed line connected by crosses); vertical vorticity (s−1 × 10) calculated from the GBVTD-estimated azimuthal wind component for the tornado (dashed line connected by dots) and the Burgers–Rott profile (dashed line connected by triangles). (b) Azimuthal wind, same as in (a); horizontal divergence calculated from the GBVTD-estimated radial wind component (s−1 × 100; dashed line connected by triangles), circulation calculated from the GBVTD-estimated azimuthal wind component (m2 s−1 × 10−3; solid line connected by squares), and radar reflectivity (dBZe; dash–dotted line connected by Xs).
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Schematic depicting the relationship between the quasi-vertical scans (very thick solid lines) by the UMass, mobile, W-band Doppler radar, through the Attica tornado, when it was mature, and the visible funnel cloud (cone-shaped, thick gray lines), opaque debris cloud (thick dark-gray lines), and semitransparent debris cloud (outer, thin light-gray lines) from 2001:44 to 2003:27 CDT. The cloud base is indicated by quasi-horizontal, thick gray lines. Ground level is depicted by thick, horizontal, solid line at the bottom. Direction of each scan (numbers plotted indicate scan number) with respect to visual features is indicated by the arrows. Scans 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 began at 2001:44, 2001:53, 2002:03, 2002:15, 2002:26, 2002:39, 2002:51, 2003:02, 2003:13, and 2003:27 CDT, respectively. At the bottom of the figure, “C” marks the location of the center of the tornado; ±rc denotes the core radius to the right (north) and left (south) of the center. The core radius is 110 m and the distance between the arrows is ∼525 m, the width of the opaque portion of the debris cloud at the ground. All features are drawn to scale. The motion of the tornado is mainly to the right (north).
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Captured frames from the boresighted video camera on the W-band radar for the 10 scans indicated of the Attica tornado. Each frame was selected when the radar antenna was oriented so that the minimum amount of the ground was visible and the tornado could be seen almost up to the cloud base. The center of the image is the approximate location where the radar beam was pointed at the time of the image; the exact location of the radar beam, which had been determined by aiming the beam at a distant ground target, was marked on a video monitor.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Quasi-vertical cross sections of W-band radar reflectivity (dBZe) on 12 May 2004 through the Attica tornado for scans 1–10, covering the period 2001:44–2003:27 CDT (times are the beginning of the first and last scan). Color scale for radar reflectivity (Z in dBZe) is shown at the lower left. Range is indicated in km, every 200 m in scans 6–9; in scan 10, range markings are shown every 100 m. To aid the reader, white line segments with double arrows indicating 400 m (300 m) are shown in each panel for scans 1–9 (10), for both vertical and horizontal scale. The bottom of each cross section is just above the ground level. The corresponding scans are depicted in Fig. 11. Scan 1 begins farthest to the right (north) of the tornado condensation funnel and debris cloud; scan 5 is through the approximate center; scan 10 is farthest to the left (south). The approximate westerly (W) direction is indicated. Also noted are the weak-echo notch (cf. Figs. 6b and 8) and a white line indicating the location of an outward bulge in the weak-echo hole/eye.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Same as in Fig. 13, but for Doppler velocity. Color scale indicates Doppler velocity (V; m s−1).
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Variations of the Doppler (quasi-horizontal) wind component (m s−1) with height (m AGL) in the Attica tornado from scans beginning from 2001:44 (scan 1) to 2003:27 CDT (scan 10), when it was mature. (a) Doppler velocity (m s−1) as a function of height (m AGL); (b) Doppler wind speed (absolute value of Doppler velocity) (m s−1) as a function of height (m AGL). The scan number for each vertical profile is shown in the insets in the upper left in each panel; upward and downward scans are indicated by “U” and “D,” respectively.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Enlarged view of the Doppler velocity in Fig. 14 (color scale to the right in m s−1; the color scale is not the same as the one in Fig. 14), for scan 5, through the approximate center of the Attica tornado. Two black circles enclose Doppler velocity couplets indicative of horizontal vortices along the outside edge of the tornado; also shown is a narrow band of flow out from the tornado, toward the radar.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Images of a tornado subsequent and to the east of the Attica, KS, tornado at approximately (a) 2012 and (b) 2022 CDT. Images are frames captured by a video (courtesy of H. Bluestein). Both views are to the southwest from locations R2 and R3, respectively, marked in Fig. 1.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Radar reflectivity imagery of the supercell that had spawned the tornado south-southeast of Medicine Lodge, KS, and near Attica, KS, on the storm scale, from the UMass, mobile, X-band, Doppler radar (UMass X-Pol) at (a) 2021:28, (b) 2038:24, and (c) 2105:46 CDT 12 May 2004 at a low-elevation angle. Hook echoes, weak-echo notches, and weak-echo holes/eyes are as noted. Range markers indicated in km and shown every 2 km. Radar was located near R3 and at two locations east of Harper on Highway 160 (cf. Fig. 1).
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Same as in Fig. 8, but at (a) 2012:25 and (b) 2024:19 CDT for the tornado just after the Attica tornado. Range markings are shown in km, every 100 m. The location of the radar was at “R2” for (a) and “R3” for (b) (cf. Fig. 1).
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Same as in Fig. 10, but at 2024:19 CDT, for the tornado just after the Attica tornado.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Same as in Fig. 11, but for just one scan, at 2024:43 CDT. The location of the radar was “R3” in Fig. 1. The two parallel, vertically oriented, thick gray lines represent the condensation funnel. Surrounding these lines is the approximate outline of the more translucent debris cloud (thin gray lines). The horizontal and slightly curved, thick, quasi-horizontal lines gray lines represent the back edge of the cloud base.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Same as in Figs. 13 and 14, but for 2024:43 CDT, through the center (cf. Fig. 21) of the tornado just after the Attica tornado. (a) Radar reflectivity and (b) Doppler velocity. The approximate southwesterly (SW) direction is indicated. Two black circles enclose Doppler velocity couplets indicative of horizontal vortices near the outside edge of the tornado.
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Vertical cross sections (height above the surface vs radius in m) of simulated motions in a tornado-like vortex for (a) small raindrops having terminal fall speeds of 2 m s−1; shaded regions indicate azimuthal speeds in excess of 40 (lighter shading) and 80 m s−1 (darker shading). Several representative streamlines are also shown. From Dowell et al. (2005, their Fig. 10a). (b) Log10 of number concentration (shading and labels at intervals of 0.2; values <−2 not shaded) of small raindrops, corresponding to (a). Arrow indicates a local minimum in number concentration outside the tornado core. (c) Same as in (a), (b), but for air motions; ordinate and abscissa represent the ratio of height to radius of maximum azimuthal wind speed in the upper-core region vs ratio of radius to the radius of the maximum azimuthal wind speed. Shading (grayscale at the bottom) represents the ratio of the azimuthal wind speed to the maximum azimuthal wind speed in the upper-core region. [From Lewellen et al. (2000, their Fig. 5)]
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Fig. A1. Regional surface map at (a) 1800 CDT (2300 UTC) 12 May 2004 and (b) 1900 CDT (0000 UTC) 12 (13) May 2004. Temperature and dewpoint temperature plotted in °F; sea level pressure plotted in hPa × 10, with the leading “10” omitted. Whole (half) wind barb denotes 5 (2.5) m s−1. Long, thin dashed line denotes approximate location of stationary front or outflow boundary; short, thicker dashed line marks approximate position of the dryline. “P28,” “LMN,” and “END” identify the station locations for Medicine Lodge, KS, Lamont, OK, and Enid, OK, respectively. The tornadic supercell described in this paper was located just east of Medicine Lodge at 1900 CDT. (Data courtesy Plymouth State College archive)
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Fig. A2. Special sounding at Lamont, OK, at 1900 CDT (0000 UTC) 12 (13) May 2004. Pressure plotted (light gray) in hPa to the left. Solid (dashed) plotted line represents temperature (dewpoint temperature) in K. Whole, half, and shaded pennant wind barbs denote 5, 2.5, and 25 m s−1, respectively. (Data and plot courtesy Plymouth State College archive)
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
Fig. A3. 500-hPa map for the United States and parts of Canada and Mexico at 1900 CDT (0000 UTC) 12 (13) May 2004. Temperature and dewpoint plotted in °C; height plotted in dam; full (half) wind barbs denote 5 (2.5) m s−1. Height contours plotted at 6 dam intervals. (Map courtesy Plymouth State College archive)
Citation: Monthly Weather Review 135, 2; 10.1175/MWR3295.1
