March 10, 2000 Supercell

I had been mentioning the severe storm threat for Friday, March 10th for over 36 hours here on the homepage by the time the event actually got underway. As luck would have it, I also worked Thursday night at KWTX-TV, filling-in for our chief weather anchor, who was under the weather, so I had an opportunity to mention it on-the-air, too. Early Friday morning, after examining the output from the ETA model (based on data from 6pm CST 3-9-2000) and from the RUC model (based on data from 6am CST Friday), I wrote:

A stalled front draped across Central Texas and an approaching vigorous upper air system will combine to produce an enhanced threat of severe storms and possible tornadoes across Central Texas today (3-10-2000). I am curtailing the maps normally available in this section in favor of emphasizing the official Storm Prediction Center and National Weather Service products. In my view, this threat is going to become focused in the central and eastern portions of Central Texas during the afternoon hours. My advice to all is to monitor official weather information sources closely. I will try to update this section with any additional information which becomes available later this morning (Friday).

I continued to monitor the weather situation through the day Friday, and at 3:30pm Friday afternoon, I wrote:

My surface analysis based on 3pm CST data shows the stalled front is along a Shreveport-Waco-north of Junction-north of Del Rio line and is nearly stationary, although in place it is oscillating north and south as outflow from morning storms north of the front competes with warmer, moist air just to the south. The main upper air jet max appears to be entering the Texas Panhandle now and divergence associated with this feature is spreading into the western parts of north central Texas near Stephenville. A mesolow (small scale surface low) is just east of San Saba and is likely to move eastward or east-northeast toward Waco over the next 2  to 3 hours. Strong to severe storms are likely to develop along the stationary boundary from east of San Saba to just west of Waco by 6pm.
Other, more isolated storms may form to the south toward Austin and San Antonio.

I should mention that I generally analyze surface pressure (PMSL) and draw isobars every two (2) millibars. The standard practice for decades has been to analyze and drawn contours every four (4) millibars when looking at the synoptic scale. It has been suggested by several authors that mesoscale analysis is facilitated by the two-millibar analysis and I find that it sometimes helps to indentify smaller features or new features which are in an incipient stage.

Here are three of the computer forecast maps which indicated the development of the general threat as early as Thursday morning. These are from the ETA model which was run on data collected at 6am on Thursday morning, 3-9-2000 and available to me by 10am the same morning.

This is the plot of surface pressure (PMSL) which was a forecast of conditions expected at noon on 3-10-2000. The computer forecast a low centered west of Austin with an inverted trough northeastward toward Tyler. The pattern of the isobars suggested a secondary push of cool or cold air coming southward through northwest Texas.

This map shows the forecast EHI for the same time (noon on Friday). Remember this is index combines both CAPE and helicity. Based solely on this map, one would have expected the most intense storms to be from around College Station eastward and southward. I also looked at the CAPE and Helicity forecast maps, and decided (mainly because of the CAPE map) that storms might develop along IH35 Friday afternoon, and become more severe as they moved toward the coast.

The icing-on-the-cake was this map, which shows the upper level wind speeds at 250mb (which is roughly 35,000 feet above sea level. A strong velocity maximum (jet max) within the jetstream flow was shown (charcoal shading) diving southeast toward west central Texas by noon Friday. These upper air features are known to produce strong divergence aloft, which permits air near the surface to rise if it is already unstable. Although the most favored areas for ascent with these features is the left front and right rear quadrants, the strength of the incoming jet max suggested that our area would experience this divergence aloft at least for a time on Friday afternoon. The position of the system also suggested that surface pressure should be falling over west central Texas, which would tend to pull the moist, warm, unstable air westward across Central Texas.

By Thursday evening, the front was stalling near the DFW area. This suggested that the threat area for severe storms might actually be closer to Central Texas by Friday afternoon, so I mentioned this on the evening weathercasts, but said that I still thought that the main threat would be southeast of our area.

By Friday morning, the threat to Central Texas was looking more pronounced. The ETA model output based on data from 6pm CST Thursday evening (not shown) shifted the areas of higher CAPE and EHI a little to the northwest. By 9am Friday morning, the situation was looking evening more threatening. Another model run by NCEP is the Rapid Update Cycle (RUC) which is similar to the ETA in terms of the equations used, but which is run every 3 hours to produce forecast output in three-hour increments out to 12 hours. The RUC basically starts with the ETA forecast for the initial time and ingests current observations and data to, in effect, update the ETA forecasts for the next 12 hours.

The images below show output from the RUC model based on data from 6am CST Friday, and valid at 6pm CST in the afternoon.

This is the RUC forecast of surface-based CAPE for late Friday afternoon. Note that there is a CAPE>4500 maximum near Victoria but that the CAPE>3000 contour extends well into our area. The sharp gradient to the west (where CAPE declines to zero) represents the dryline.

Well, there was still the question of whether storms would develop at all. The K Index is often considered a good measure of whether the atmosphere will initiate convection. You can see that there was a K>32 contour right over the area from Burnet to northwest of Waco, and that value was indicative of the likelihood that convection would develop in that area by 6pm. The white lines represent sea level pressure in millibars (mb) with contours every 4mb. For analyzing severe storm potential, I like to see the contours every 2mb, so the next maps shows that.

This map shows two parameters: the PMSL drawn for every 2 millibars (white) and the divergence of the surface wind (actually the wind at B015, a model level near the surface) in red. You'll note that the red contours in Central Texas are negative. That indicates
strong convergence of the surface flow in this area (because "negative" divergence equals convergence). Notice how the convergence is wrapped around the north side of the surface low (the area within the 1008mb white contour). This indicated to me that the surface flow east and northeast this low center would be "backed" ... that is, it would remain more southeasterly or even easterly. The impact of a "backed" flow is to enhance storm-relative helicity. Also, note the packing of the isobars across north Texas and Oklahoma; I took this to be evidence of cool (or cold) air pressing southward, which would tend to force the surface low to move east rather than northeast.
Further evidence of that could be found in the plot of forecast surface (2m) temperatures, which is the map below [temperature is in degrees celsius (C)], with contours every 2 degrees. A rough conversion of the data is that the forecast temperature at Abilene was 20 deg C (68 deg F) while the forecast temperature at Hondo was 30 deg C (86 deg F).

Well, I looked at a lot of additional output from the 6am CST RUC but these maps pretty well tell the story. The threat area had definitely moved westward into Central Texas! By 9am Friday, it was time to monitor the evolution of the system and compare what was actually being reported to what the models were forecasting.

Strong thunderstorms developed across northern Texas and southern Oklahoma during the morning hours, mostly in the area north of the stalled front. These storms were what are termed "elevated", meaning that they are drawing on moisture and instability that is not "surface-based". In other words, the moisture and instability feeding these storms was aloft, above the cool air north of the front. Put another way, the front was shallow. Despite being "elevated", these storms produced an impact at the surface: heavy rain and hail feel from some of them, and they all produced cool outflows which served to reinforce the temperature gradient across the stalled front, which actually began sinking southward again, fueled by the renewed cooling at the surface.

By 12 noon CST, the surface map looked like this.

There are many different ways to monitor the evolution of a system like this one: one can use surface observations, or radar, or satellite imagery, or profiles of the winds aloft (generated by either vertical profilers or by the nexrad radars).  The best approach is probably to use all of these. Of particular value on this date was the high-resolution visible imagery generated from the geostationary satellites (GOES), which provide frequent images showing objects as small as .6 mile in diameter (i.e. 1km). The next images is a high-resolution visible image taken at 2010UTC (2:10pm CST) and centered on Waco, Texas. At this time, the GOES system had been placed in rapid-scan mode and was providing a new image every 5 minutes.

On the image I have annotated four weather features. The green line represents the outflow boundary/stalled front, which was advancing slowly southward on the west end but more rapidly on the east end, where strong storms shortly after noon had pushed cool air southward. The yellow line represents the dryline which was advancing slowly west to east. The blue line represents a push of cooler air behind which the winds shifted to the north and increased in velocity. The red "L" represents a low pressure area which I called a mesolow on postings on Friday.  A mesolow is a small scale low pressure center which frequently forms at the junction of an outflow boundary, dryline line and cold front. The junction of those features is called a "triple-point".   Clicking here will take you to the same image without the annotations, for a better view of the cloud features associated with these features.

Also notice these other important features: (a) the relatively weak storms forming in McCulloch and Mason counties along the dryline, with the tops of these storms being blown to the east-southeast, giving a hint that the storm-relative winds at upper levels are from the west-northwest. This would be a very important indicator that conditions conducive to supercell storms were evolving. And (b) the relatively cloud-free area just east of the mesolow in at the triple-point. This clearing permitted increased solar radiation to reach the surface, destabilizing the atmosphere rapidly in that area.

Remember the EHI parameter? Well, it's not only a forecast parameter, some analysis programs can compute an estimate of EHI in realtime. The Storm Prediction Center does just that, and the map below shows the 2pm CST computed EHI considering only the change in wind speed and direction (the helicity part of the equation) below 1 kilometer above the ground (that's roughly from the surface to 3,000 feet AGL). This is a particularly important parameter for tornado prediction, because recent studies suggest that it is the increase in velocity and directional turning in this layer which most affects whether a tornado forms. Here's the analysis:

Note the >3 maximum bullseye which is centered just about where I placed the meso-low in the analysis shown on the satellite image.

Another parameter I like to use for situations where convection appears likely is moisture convergence. Remember the RUC forecast map above which showed forecast divergence (convergence) of the wind. It is possible to perform this type of analysis with respect to both wind and the moisture it contains. This is called moisture convergence, and again, various computer programs can perform this analysis. The ones I've chosen so show you come from the Center for the Analysis and Prediction of Storms (CAPS) site at the University of Oklahoma. I have several of them to show you, but in the interest of saving loading time, I'll only put two in-line and let you "click" to see the others.

This is a zoomed portion of the 2pm (2000UTC) moisture convergence analysis from CAPS. The full image is available here. The yellow identifies the area of strongest moisture convergence to be over western McLennan County. A tongue of enhanced moisture convergence extends from near Temple northwest to near Meridian. A single hourly moisture convergence analysis may or may not be of significance, but when you see the pattern persists for several hours in the same area, Mother Nature is waving a red flag at you! Three hours later, the moisture convergence analysis looked like this:
Note that the yellow area (strongest moisture convergence) has persisted in the same general area of western McLennan, eastern Coryell and northern Bell counties! The analysis is telling us that this zone must be watched because any storm forming or moving into this area has the potential to become severe very quickly. Of course, at the time of this analysis (5pm CST), there was a severe storm in western Coryell County and it was moving toward this zone. The links below lead to the full images of the moisture convergence analyses at 3pm and 4pm CST.
     Click for the 3pm CST moisture convergence full image.
      Click for the 4pm CST moisture convergence full image.

Here is a zoomed portion of my 3pm CST surface analysis map. It's a little messy because it was drawn in some haste as the storms became more active. The handwritten numbers below and to the right of some stations show the 2-hour pressure change in millibars and tenths (i.e. -25 would be a 2.5mb decrease in surface pressure over two hours time). There is a zone of >2mb pressure falls over this period stretching from Waco to Lufkin. When other factors are also present (as they were in this case), the zone of maximum pressure falls may also give an indication of where storms will intensify.

Compare this map with the RUC surface forecast valid at 6pm CST (shown above) and you'll see that the RUC did a pretty good job on surface features. That's often an indication (but no guarantee) that the model has a good handle on the other features we were looking at, too. Here's a link to a nice, clean version of the map without any analysis so that you can read the station data more clearly.   Click here to see the 3pm surface map without analysis.

Well, when you get to this point in the day and storms are developing, the very best forecasting (actually, nowcasting) tool is probably the weather radar. So, let's finish this up with a look at selected radar images. Unfortunately, I wasn't able to obtain any images from the closest nexrad site (the DoD radar at Granger, controlled from Fort Hood) but you can see what happened using the Fort Worth NWS nexrad, too.

Here's the first image, from the KFWS (Fort Worth) site at 4:33pm CST:

The radar shows a discrete cell in western Coryell County (just above the KGRK/KILE identifiers), and a multi-cell cluster of storms over Hamilton and Bosque counties, extending north to a radar thin-line feature which stretches from KFWS (the radar site) northeastward to the Red River. The thin-line feature is probably the reinforcing cold front. The discrete cell in Coryell County is beginning to show features commonly associated with supercell storms, which include the downwind "V" pattern of the reflectivity contours and the the strong reflectivity gradient on the upwind (relative to upper flow) flank.

Here's the second image from the KFWS radar at 5:03pm CST:

The radar shows the multi-cell features which had been in Hamilton and Bosque counties accelerating southeastward and becoming more linear. The discrete cell which had been in western Coryell County had moved to near Gatesville and appears to be weakening. However, note that the radar beam must pass through the intense convection over Bosque County to reach the cell near Gatesville, so there may be some attenuation of signal. The shape of the cell in Coryell County is still suggestive of a storm with a single, strong updraft. Finally, it appears that the southeastward moving line may overtake the discrete cell.

Here's the 5:33pm CST image from KFWS:

At this point, it appears that the line of storms has overtaken the Coryell County cell. However, note the intense (>65dBZ) portion of the echo in central Coryell County. Hail as large as 4.5 inches diameter was falling from this portion of the storm. A very close examination of this image reveals that the cell is actually still discrete, not a part of the line. At this time, we were having trouble with our radar at KWTX-TV because our radar site (near Moody, in the southern tip of McLennan County) was receiving rain, causing attenuation due to a wet radome, and because we, too, were looking through intense convection to reach the storm in Coryell County.

Here's the 6:03pm CST image from KFWS:

With this image is becomes obvious that the discrete cell in Coryell County was not absorbed by the line of storms. In fact, it is obvious that there is yet another strong storm developing over the western area of Coryell County. However, at this time, the cell just crossing from Coryell to Bell was the primary concern, because it had developed a well-defined hook echo and was triggering the "MESO" algorithm on the KGRK radar (not shown). The hook was first identified about 5:55pm on our KWTX Doppler10 radar, and persisted for almost 30 minutes. A couple of minutes after this radar image, we were able to use the KWTX Killeen Skycam to visibly observe a well-defined wall cloud with rotation about 10 miles north of Killeen, over the Fort Hood military reservation. We tracked the wall cloud witht the skycam for about 5 minutes, then observed cyclonic rotation in the cloudbase even after the wall cloud dissipated. On two separate occasions, we observed "clear slots" wrapping cyclonically into the cloudbase rotation after the wall cloud had dissipated. At 6:23pm CST, NWS Fort Worth issued a tornado warning for Bell County, citing a severe thundertsorm with "tornadic circulation 6 miles west of Morgans Point Resort moving southeast".

Here's the 6:32pm CST image from KFWS:

The cell approaching Belton and Temple (central Bell County) looks considerably more "outflowish", which would indicate a diminished tornado threat. New cells (aligned with the southeastward moving line which crossed Waco) are developing southeast of the Bell County storm, possibly impacting the quality of air available to the discrete storm.

And here's the last radar image for this event, from KFWS at 6:52pm CST:

Interestingly, the Bell County storm refuses to die, and is again showing some characteristics of a supercell storm, including the lower reflectivity echoes stretching northwest from the west flank of the storm, which appear to be some sort of flanking line feature. This storm eventually departed Bell County into Milam County with no further tornadic activity. A few reports of hail and damaging winds were received from southeast Bell County as the storm moved southeastward. The line of storms (seen here from Limestone County to eastern Williamson County) moved rapidly southeast and produced derecho-like events all the way to near Houston.

And to sort of tie this all back to a weather map, click here to see the 6pm CST surface analysis.

    ETA model forecast images were obtained via the courtesy of Gilbert Sebenste and the Storm Machine
    RUC output was obtained in gridded format from the NWS OSO server and processed on PC-Gridds.
    Radar plots, maps and analysis from CAPS was obtained from the CAPS server and from the archived files at CAPS.

    CAPS is The Center for Analysis and Prediction of Storms (CAPS), located at the University of Oklahoma in Norman, and is a National Science  Foundation Science and Technology Center whose mission is to demonstrate the practicability of small-scale numerical weather prediction with an emphasis on deep convection.

  Back to the main homepage!

This page was last updated on 4-10-2000.

If you have comments or suggestions, email me at

This page created with Netscape Navigator Gold