Figure 1

On 18 July 2001 severe thunderstorms developed over the north central plains of the United States. The synoptic pattern was dominated by a large ridge over the south central plains, however an upper jet north of the ridge allowed sufficient shear to support supercell thunderstorms over the Dakotas. Figure 1 shows the CIMSS GOES derived wind analysis from 20:45 UTC on 18 July. Wind barbs are in knots and color coded by pressure level where black is 100-250 mb, cyan is 251-350 mb, yellow is 351-500 mb. The analysis utilizes a sequence of three water vapor images to compute winds using the NOGAPS model as the first guess field. 300 mb Isotachs are in red (in m/s). The image shows a jet streak over Wyoming with diffluence over the Dakotas.  Click on images to enlarge or start loops

Figure 2

Figure 3

Figure 4

The evening sounding taken at 00:00 UTC 19 July (Figure 2) at Bismarck, ND shows that CAPE values had increased to 5000 J/Kg, the LCL/LFC was relatively low (at a height of 835 mb), and that the winds aloft had increased with the arrival of the upper jet. The 00:00 UTC 19 July MesoEta analysis of surface CAPE, MSLP and winds (Figure 3) shows CAPE values around 5000 J/kg in eastern North Dakota. A trough is situated along the western edge of the high CAPE axis.  Converging surface flow can be seen along this trough and visible satellite imagery (Figure 4) shows a line of towering cumulus developing along this line. The NESDIS FPDT single field of view CAPE product in (Figure 5) shows the instability may have been even greater than the MesoEta analysis, but the location seems to correspond well. 
 

Figure 5

Figure 6

Figure 7

The visible satellite loop (Figure 6) shows storms developing along the boundary mentioned above beginning about 21:30 UTC. A series of interactions may seen between several storms along this line. Between 22:30 and 22:46 UTC one cell along the line (Storm A) rapidly develops a near vertical wall of towers along its west side (brightly lit by afternoon sunshine). In the period from 22:46 to 23:00 a back-sheared anvil expands rapidly westward. It is during this time period that the first tornado is reported. A smaller, shorter-lived storm, (Storm B) develops to its immediate south and an outflow from that cell intersects Storm A between 23:00 and 23:23 UTC. An apparent storm merger seems to occur shortly thereafter between the two storms. During this time, storm chaser Paul Sirvatka observed a large tornado. Storm A appears to split sometime between 23:46 and 00:00 UTC. Finally, a third storm (Storm C) forms near Jamestown, ND between 23:30 and 00:00 UTC. It produced large hail and later (1:53 UTC) a tornado. 

Figure 7 is a loop of 0.5° base reflectivity from the KMVX radar in Grand Forks, ND spanning the period from 22:58
through 00:28 UTC. NLDN cloud-to-ground lightning data are overlaid. Notice the merger between Storm A and Storm B during the period 23:13 and 23:30 UTC. This is the merger described above (Figure 6). Though the range to the storms is too great to show the outflow moving northward from Storm B into Storm A, the radar data confirm that a storm merger took place. As this interaction took place, Storm A became dominated by positive cloud-to-ground lightning strikes. Shortly after this interaction an outflow boundary appeared on the west side of storm A.