Eric Martello(1)
National Weather Service Office
Goodland, Kansas



On August 21, 1997, two tornadic supercells formed in southwest Nebraska, merged, then moved rapidly into northwest Kansas as a bow echo system (Figure 1). Prior to the merger, the two storms produced 4.5 cm (1.75 in) diameter hail just before crossing the Kansas/Nebraska state line. After the storms evolved into the bow echo, wind damage became more pronounced and widespread as the system moved rapidly southeast across portions of northwest Kansas. Although there were some differences during the mesocyclonic stage, the event was similar to one that occurred near Lahoma, Oklahoma in 1994 (Conway1994). This study will examine the environmental conditions, radar signatures, and structural damage associated with the event.

Figure 1. Track of two supercells and resultants bow echo of August 21, 1997.


Area soundings at 1200 UTC 21 August, showed that a surface-based inversion existed over western Kansas with surface temperatures ranging between 16-19°C and 850 mb temperatures in the 26-28°C range. At this time, Convective Available Potential Energy (CAPE) values of 1000-1500 J kg-1 existed over the region. A northwest-southeast oriented 500 mb baroclinic zone existed over central Nebraska and Kansas between a low over the upper Midwest and a ridge situated over the central Rocky Mountains, resulting in a northwest flow regime over western Kansas. 1200 UTC Lifted Indices (LI) at 1000-500mb ranged between -2°C and -4°C over the region (Figure 2). By 0000 UTC 22 August, ample surface heating had broken the surface-based inversion with LI values approaching -10°C (Figure 3). Mid-level (700-500mb) lapse rates over Goodland, Kansas, were 7.7°C km-1 based on a modified sounding taken from Dodge City, Kansas, at 0000 UTC 22 August.

Figure 2. Surface-based lifted indices (°C) 1200 UTC 21 August 1997.

Figure 3. Surface-based lifted indices (°C) 0000 UTC 22 August 1997.

The modified sounding for Goodland, based on a sounding from Dodge City, Kansas, using Eta temperatures and dew points up to 500 mb (Figure 4) at the same time, indicated advection of dry air from the southwest at 750 mb. It also showed that a dry, westerly flow off the Rockies existed at 700-600mb with the potential for strong evaporative cooling in this layer increasing the potential for large hail and downbursts. A shallow, moist layer existed between 850-800 mb. A south wind at these levels backed to the southeast at the surface. CAPE values increased to between 4000 and 5000 J kg-1 by late afternoon with 32°C and 20°C surface temperatures and dew points, respectively, at 0000 UTC 22 August. In this case, the system developed in existing high CAPES, whereas, the 1994 Lahoma system advected into a steep CAPE gradient which increased storm-relative (SR) inflow into the system and resulted in a change in storm structure as the system moved south-southeast (Conway 1994).

Figure 4. Modified sounding for Goodland 0000 UTC 22 August 1997, showing Convective Available Potential Energy (CAPE) area.

Johns (1993) states that high CAPES are an important ingredient for long-lived warm season bow echoes in the absence of strong dynamic forcing aloft. Once the 1.6°C cap shown in the modified sounding was broken, thunderstorms developed rapidly. A storm-type derivation chart (Brooks and Doswell 1994), shown as Figure 5, was used in conjunction with the modified sounding for Goodland to estimate the probable storm-type of the activity. The chart uses mixing ratios and helicity values to determine types of severe weather associated with thunderstorms. The location of Goodland's parameters for the afternoon on the chart indicated that straight-line-wind-type storms were a strong possibility. The potential for tornadoes was significant as well, and tornadoes did occur during the mesocyclone stages of the two supercells.

Figure 5. Mixing ratio and helicity values for Goodland 0000UTC 22 August 1997 plotted on Brooks and Doswell's (1994) storm-type chart.

However, no horizontal vorticity-induced tornadoes were reported during the bow echo stage. Although mesocyclones may develop along the "bookends" of a bow-echo (Johns 1993), only damaging straight-line winds occurred with this event.


The mean sea level (MSL) pressure pattern at 0000 UTC indicated a lee trough from the Nebraska panhandle to the eastern plains of Colorado. Dew points across the lee trough ranged from 12-15°C in eastern Colorado to 19-22°C in eastern portions of northwest Kansas and southwest Nebraska (Figure 6). An east-west quasi-stationary front existed from west-central Nebraska southeast to near Concordia, Kansas. Johns (1993) found that weak wind-extreme instability events tend to occur along and to the south of stationary fronts, similar to the situation in place over northwest Kansas. A surface to 850 mb Theta-E ridge (Figure 7) existed from Texas into western Nebraska with a developing 25 to 35 kt southerly low-level jet (LLJ) situated east of the surface trough. At 2330 UTC, two isolated supercells formed in southwest Nebraska east of this trough, which is a typical location for convective initiation (Kriehn 1992). Winds of 25 to 35 kts were noted in the 600mb-700mb layer at this time (Figure 11). Winds of this magnitude are strong enough to produce tornadogenesis within a storm-relative (SR) helicity field of 150-175 m2s-2 (Brooks and Doswell 1994). In the Lahoma ,Oklahoma event, SR mid-level winds of 15 to 20 kts (Conway 1994) were not enough to generate and stretch the low-level horizontal vorticity in the mesocyclonic stage of the supercell, as happened in the northwest Kansas case. However, both events transformed into bow echoes after moving into higher low-level moisture regimes. Johns (1993) states that in extreme instability conditions, weak mid-level winds are common in warm season bow echo events. In addition, he suggests that optimum condition for bow echo maintenance consists of strong vertical wind shear in the lowest 2.5 km AGL with unidirectional winds in the mid-levels. Wind profiles the evening of August 21 correlated well with Johns' observations.

Figure 6. Surface observations, analyzed surface pressures, fronts, and dew point temperatures for 0000 UTC 22 August 1997.

Once boundary layer winds decoupled from the surface, the LLJ shifted slightly east by mid-evening. Such a shift is primarily due to thermal wind relationships related to differential cooling associated with the east-west sloping terrain (Bluestein 1984). Similar to a case documented by Kriehn (1992), these storms moved into a region of lower levels of free convection (LFC) where deeper low level moisture existed. This is similar to the Lahoma case where both events occurred in environmental surface mixing ratios around 17 g kg1. Less lift is needed to initiate convection where lower LFCs exist. By 0300 UTC 22 August, this system behaved similarly to Kriehn's observations as outflow produced by the full-fledged bow echo system initiated new convection along its front flank.

The bow echo propagated south-southeast into the nose of the Theta-E ridge over west-central Kansas (Figure 7), then into northwest Oklahoma by 0900 UTC 22 August. Strong SR inflow from a 30 to 40 kt LLJ over western Oklahoma sustained the bow echo through the night. By 1200 UTC 22 August, the storms dissipated as SR inflow decreased and the nocturnal LLJ weakened.

Figure 7. Surface-850 mb layer Theta-E analysis for 0000 UTC 22 August 1997.


After the two rotating supercells formed over southwest Nebraska, the larger northern cell turned right of the mean flow. By 0049 UTC 22 August, it began to merge with the slower and smaller supercell to the south over Chase County, Nebraska (Figure 8). Storm Top Divergence was estimated by the Goodland WSR-88D radar to be 184 knots as the two supercells merged between Imperial and McCook, Nebraska. Divergent flow aloft of this magnitude is strong enough to produce hail up to baseball size (WSR-88D Operational Support Facility 1993). Spotter reports from Dundy County in southwest Nebraska indicated hail 4.5 cm in diameter fell near locations of the maximum reflectivity on Goodland WSR-88D's mid-level (24-33Kft) LRM product (Figure 9). Both supercells produced tornadoes around this time (0045 UTC) and the Goodland WSR-88D radar showed the associated cyclonic rotation signatures on the 0.5° base velocity product (Figure 10).

Figure 8. Goodland WSR-88D 0.5° base reflectivity product for 0049 UTC 22 August 1997.

Figure 9. Goodland WSR-88D mid-level layer reflectivity maximum product (LRM, 24-33 kft) for 0142 UTC 22, August 1997

Figure 10. Goodland WSR-88D 0.5° storm-relative velocity product (SRM) for 0055 UTC 22 August 1997.

Figure 11. Goodland WSR-88D velocity azimuth display (VAD) wind profile product for 0005 UTC 22, August 1997, showing wind direction and velocities through the vertical in kft MSL.

As the storms merged, they moved into a more unstable thermodynamic environment. Evaporative cooling in response to dry mid levels helped to generate strong baroclinic vorticity within the storm complex. Once this occurred, the downdraft strength increased and formed the familiar spearhead shape indicative of an intensifying rear inflow jet (RIJ) (Przybylinski and Gery 1993). Between 0130 UTC and 0230 UTC, the RIJ and associated precipitation helped to undercut the low-level updraft which occluded the storm and eliminated the mid-level mesocyclone, similar to the Lahoma case (Conway 1994).

Two major differences arise between this event and the 1994 Lahoma, Oklahoma event during the initial mesocyclonic stage. Only one supercell existed prior to transition in the Lahoma case where two merging supercells existed in the northwest Kansas case. Furthermore, the Lahoma storm did not produce a tornado in its early mesocyclone stage whereas in this case both supercells produced tornadoes. Stronger mid-level winds in the northwest Kansas case most likely attributed to the tornadogenesis that occurred.


By 0330 UTC, the LLJ in the boundary to 850 mb layer had developed. The Red Willow, Nebraska profiler showed this feature strengthened to 40 knots. This LLJ helped to feed in high Theta-E air into the broadening thunderstorm complex. The convection then moved into an environment consisting of a lower LFC, higher CAPE, and a more linear wind profile in northeast portions of Goodland's CWA, similar to the later stages of Conway's (1994) observation of the Lahoma, Oklahoma event.

A well-defined weak-echo region can also be seen along with an outflow boundary from the southwest flank of the storm extending west to Goodland. Furthermore, a strong reflectivity gradient existed along the front edge of the thunderstorm complex (Przybylinski and Gery's 1993). In addition, the maximum echo top was over or ahead of the strong front flank low-level gradient (Figure 12). Fujita (1978) described this characteristic of a mature bow echo.

Figure 12. Goodland WSR-88D Echo Tops (ET) and base reflectivity product for 0329 UTC 22 August 1997 showing radar indicated storm tops during bow echo stage.

The storms continued to move southeast, reaching Hays, Kansas (HYS) around 0430 UTC. Base velocity estimates at that time (Figure 13) indicated the potential for 80 knot winds mixing to the surface. A strong rear inflow notch (RIN) can be seen in the corresponding reflectivity (Figure 14) as mid-level evaporative cooling intensified the downdraft and produced an outflow surge along the front edge of the thunderstorm. However, these storms were moving nearly perpendicular (70-80°) to the radial of Goodland's WSR-88D electromagnetic beam. Since the WSR-88D estimates velocity best when scatterers move parallel to the radial of the beam, this likely led to an underestimation of mid-level winds. In addition, velocity estimates were probably degraded due a decrease of the beam's power density through volume averaging spatially as one moves away from the RDA. Przybylinski and Gery (1993) suggest that such limitations on the WSR-88D can be supplemented by a good spotter network to verify severity of ongoing weather.

Figure 13. Goodland WSR-88D 0.5° base velocity product for 0439 UTC 22 August 1997, during bow echo stage.

Figure 14. Goodland WSR-88D 0.5° base reflectivity product for 0439 UTC 22 August 1997, during bow echo stage.


During the period of transition (0130-0230 UTC), the convective storm structure observed on the Goodland WSR-88D reflectivity loop began to resemble a pattern that implies damaging winds and downbursts, as described by Fujita (1978). At 0153 UTC, the Automated Surface Observation System (ASOS) in McCook, Nebraska took a special observation which recorded west winds sustained at 26 kts gusting to 33 kts. Convective activity at this time was approximately 80 km to the west of the McCook ASOS. Southeast Dundy County experienced wind gusts in excess of 58 kts and hail 4.5 cm in diameter ten minutes prior to the McCook observation (Figure 15). At that time, the merging supercells still had their mesocyclonic characteristics with tornadoes and large hail being the main concern of the operational staff.

The storms began to surge into northwest Kansas by 0200 UTC. A bow echo took shape at that time as the RFD was intensifying. The extent of damaging winds was approximately 85 km wide, extending from Rawlins County, Kansas to Norton County, Kansas. Wind gusts greater than 58 kts were reported by several spotters in this area (Figure 15). Downed power poles, uprooted trees, blown out windows, and siding ripped off the sides of homes were reported near Dresden, Kansas. Johns (1993) states that bow echo induced downbursts account for the majority of casualties and damage from convectively-induced non-tornadic winds.

The RIN and associated weak-echo region (WER) continued to show further intensification as the system moved into Graham and Sheridan Counties of northwest Kansas after 0300 UTC. The system reached the Hill City ASOS site (HLC) at 0331 UTC with sustained winds up to 59 kts and gusts to 68 kts.

The intense rain kept visibilities below one statute mile (1.6 km) from 0325 UTC until around 0355 UTC. Sheridan County to the west reported similar conditions resulting in large tree limbs being snapped off, campers being flipped, and electricity being knocked out at Sheridan Lake near Hoxie, Kansas (Figure 15). The storms produced a 50-knot gust in Hays, Kansas at 0436 UTC.

The bow echo continued south-southeast into southwest Oklahoma. At 1400 UTC, the GOES 8 Infrared (12µ) satellite imagery showed remnants of this system between Lawton, Oklahoma, and Wichita Falls, Texas. Storm damage from this system emphasized that warm season, mesoscale induced bow echoes can be as intense as dynamic-type bow echos, that occur more often in fall and spring.

Figure 15. Storm Data complied in local standard time (LST) for areas in Goodland's County Warning Area (CWA).


Changing atmospheric environments and subsequent effects on thunderstorm structure were instrumental in the evolution of the August 21, 1997, supercell merger into a bow echo system. Potential moisture availability, strength of LLJ feeding the convection, and SR mid-level winds, thermodynamic instability (CAPE, LI, etc.), and vertical/horizontal wind shear all played vital roles in how the convection formed and behaved. A thunderstorm's internal dynamics can change the ambient environment in which it resides in (Fujita 1978), which is what occurred with the two supercells as they merged together north of the Kansas/Nebraska state line. Strong SR low level inflow combined with high CAPES induced precipitation loading at the top of the updraft once the LLJ developed after sunset and fed into the system. The consequential precipitation cascade evaporating in the dry mid levels created a strong RFD descending from the mid levels. The resulting RIJ surged out of the storm's leading edge cutting off inflow into the mesocyclone and developing the wedge-shaped signature.

The Goodland WSR-88D provided vital clues to the convective metamorphosis. The Echo Top radar product showed a migration to the front edge of the convection. Reflectivity products in the lowest two elevation scans gave early signs of a "wedge" developing as the storms moved into the nose of a 30 to 40 knot LLJ. The "classic hook" of the merging storms soon disappeared as the low level flow into the mesocyclones was cut off. Once into a higher CAPE and moisture environment, the "wedge" in reflectivity sharpened and showed a weak-echo region (WER) where the strong RFD transferred momentum down to the surface.

The Goodland WSR-88D's ability to estimate velocities along the radial of the beam indicated the stronger winds associated with the RIJ. Due to larger volume averaging of the radar's electromagnetic beam, velocity measurements of storm-outflow were likely underestimated. Furthermore, the eventual perpendicular orientation to the Doppler radar's beam added to estimations misrepresenting actual wind. Even so, WSR-88D velocity products were instrumental in estimating downdraft potential.

The RIJ cutoff of SR low level inflow into the merging storms dissipated the mesocyclone, leading to a more multicellular structure. Consequently, after the RIN developed on the base reflectivity scan, straight-line wind damage began to occur. A mesoscale observation network (mesonet) combined with a well-trained spotter network were invaluable when confirming signatures on radar. This convective event demonstrated that warm season bow echoes can be just as intense as dynamic-type fall and spring systems. Storm damage from this warm season event documents how bow echoes, in general, can produce destructive winds and can pose a serious threat.


Thanks to Timothy O. Mathewson , National Weather Service Office (NWSO) Goodland, Kansas for his efforts in obtaining environmental data through GEMPAK, to Joy Hayden, Administrative Assistant, NWSO Goodland, Kansas for performing a literature search on bow echoes, and to J. William Conway, National Severe Storms Lab (NSSL) for supplying me with a copy of his 1994 research on the Lahoma, Oklahoma bow echo event.


Bluestein, H.B., 1984: Fronts and Jet Streaks: A Theoretical Perspective. Mesoscale Meteorology and Forecasting, Peter S. Ray, Ed., American Meteorological Society, Boston, 173-215.

Brooks, H.E., and C.A. Doswell III, 1994: On the Environments of Tornadic and Non tornadic Mesocyclones. Wea. Forecasting, 9, 613-618.

Conway, J.W., H.E. Brooks, and K.D Hondl, 1996: The 17 August 1994 Lahoma, OK Supercell: Issues of Tornadogenesis and Bow Echo Formation. Preprints, 18th Conf. Severe Local Storms, San Francisco, AMS (Boston), 52-55.

Fujita, T.T., 1978 Manual of Downbursts Identification for Project Nimrod. SMRP Research Paper-156, Univ. Of Chicago, 42-44.

Johns, R.H., 1993: Meteorological Conditions Associated with Bow Echo Development in Convective Storms. Wea. Forecasting, 8, 294-299.

Kriehn, T.E., 1993: The June 16,1992 Severe Weather Outbreak...A Case Exhibiting Supercell and Bow Echo/Type II Derecho Development. Central Region Applied Research Papers 10. DOC, NOAA, NWS Central Region, Scientific Services Division, Kansas City, MO, 127-138.

Operational Support Facility (OSF), 1993: Student Guide for WSR-88D Training Course. Topic 5 Lesson 2 Convective Storms: Structure, Evolution, and Radar Depiction. Operations Training Branch, Operational Support Facility, DOC, NOAA, National Weather Service, Norman, OK, 118pp.

Przybylinski, R.W., and Gery, William J., 1993: The Reliability of The Bow Echo as an Important Severe Weather Signature. Central Region Applied Research Papers, 9. DOC, NOAA, NWS Central Region, Scientific Services Division, Kansas City, MO, 270-273.


1. Current Affiliation, NWS Jackson, Mississippi. is the U.S. government's official web portal to all federal, state and local government web resources and services.