SR/SSD 98-10


Technical Attachment


Henry Steigerwaldt

Science and Operations Officer

NWSO Nashville, Tennessee

1. Introduction

During the evening of February 3 and into the morning hours of February 4, 1998, a snowstorm struck the east half of middle Tennessee, producing up to a foot or more of snow in some counties. The heavy wet snow knocked down tree limbs and power lines, causing power outages to an estimated 90 thousand people, and stranding around 600 people on Interstate 40 in Putnam County for up to 18 hours. Some people were still without electricity one week after the storm.

During the evening of February 3, precipitation changed from rain to snow over middle Tennessee. What was unusual about the changeover was that it occurred some 6 to 9 hours earlier than forecasters had expected. It is believed that dynamic cooling was the primary mechanism that resulted in cooling the lower and middle levels of the atmosphere, allowing the rain to change to snow much earlier than anyone expected.

The subject of dynamic cooling seems to become a topic of discussion among meteorologists only when a change from rain to snow occurs unexpectedly. Just what is meant by dynamic cooling anyway? This paper will briefly discuss the role of dynamic cooling and events leading up to and after the precipitation change that occurred in middle Tennessee. The discussion and figures that follow will be used to support the belief that dynamic cooling was responsible for the change from rain to snow. This storm had some similarities to the snowstorm that struck mainly south-central Mississippi and west-central Alabama on December 14, 1997, in which dynamic cooling also appears to have played a role.

2. Temperature Change

There are a number of factors that can contribute to a change in temperature at a particular location. These factors include terrestrial and solar radiation, latent heating and evaporative cooling, sensible heat exchanges in the planetary boundary layer, and of course, advection and vertical motion. Meteorologists generally devote much of their attention to the last two factors, namely advection and vertical motion.

It is common to think of advection as being horizontal in nature. Warmer or cooler air flows from upstream to downstream areas. In reality however, advection may not be horizontal at all. As long as a parcel is not saturated, it is bound to a particular isentropic surface, and cools or warms at a fixed rate. If a diabatic (non-adiabatic) process occurs, such as latent heating or evaporation, a parcel is no longer bound to a particular isentrope. It will cross isentropic surfaces as it moves along. The parcel's temperature will still decrease or increase, but at a slower rate.

Most textbooks consulted had no reference to dynamic cooling or dynamic heating. One textbook however (Geddes 1939) mentioned "dynamical cooling," and that was in the context of "cooling by expansion." Dynamic cooling is therefore the cooling that results from decreasing pressure. Similarly, dynamic heating is the heating that results from increasing pressure. Because the pressure gradient is much stronger in the vertical than in the horizontal, "dynamic" changes in temperature due to expansion or compression are more likely to occur from vertical motion than from horizontal motion.

Factors other than advection and vertical motion also contribute in part or in total to a change in temperature. However, when considering a location aloft, if the change in temperature cannot be explained by pure advection, then vertical motion is probably the major factor for the temperature change.

The occurrence of dynamic cooling or heating is most often noticeable when a closed isotherm appears on an analysis, but was not depicted 12 hours earlier. The location of the colder (warmer) temperature is usually near a closed low (high) pressure center where vertical motion is at a maximum.

3. Conditions Before the Changeover from Rain to Snow

Figure 1 shows the GOES-8 IR imagery for 1715 UTC February 3, 1998. The surface low was located over the Florida Panhandle south of the Alabama and Georgia state line, with a 500 mb closed low estimated to be nearly overhead. Although somewhat fragmented, a deformation zone cloud system associated with the closed upper level low extended from Alabama and Mississippi, north across western and middle Tennessee, western Kentucky, and into Illinois and Indiana. A vorticity comma cloud system was located to the east of this cloud system.

Before the change from rain to snow occurred, the heaviest rainfall in middle Tennessee had fallen over the southeast part of middle Tennessee, where 2 to 3 inches of rain was common. The moderate to heavy rain was associated with both the deformation zone and vorticity comma cloud systems. Recognizing deformation zone cloud systems is important since they are commonly associated with moderate to heavy precipitation (Steigerwaldt 1986). Just such a cloud system was also present during the snowstorm that struck mainly south-central Mississippi and west-central Alabama on December 14, 1997, which produced significant snowfall.

Leading up to the eastern middle Tennessee event, the models were forecasting the surface low to move generally northeast into southern Georgia and then east into the Atlantic, before taking a turn to the northeast. The 500 mb low was forecast to move north across Alabama, and then shift east across Georgia and into the Atlantic before it too would begin a turn to the northeast.

The models, particularly the Eta and NGM, grossly underforecast the strength of the storm system, both surface and aloft. Because of the more intense storm system, it seems reasonable to assume that stronger upward vertical motion was not accounted for in the models. The NGM was the worst of the three models that were used (Eta, NGM, AVN) in verifying the storm's strength from the surface through 500 mb, however, its forecast position of the 850 mb zero degree isotherm gave the best hint that colder air might drop south, primarily into the Eastern Highland Rim and Cumberland Plateau areas (Fig. 2) during the night.

Based on the NGM forecast of the 850 mb zero degree isotherm dropping southward during the night, it was believed that the rain over middle Tennessee would become mixed with snow in the Nashville area (northern middle Tennessee) sometime after midnight, and change to snow over the higher terrain of the northern Cumberland Plateau after midnight. The Eta and AVN models also showed some cooling at 850 mb, but not as much.

4. Conditions During and After the Changeover

The first sign that the rain was changing to snow much faster than anyone anticipated (6 to 9 hours faster, in fact) was when a weathercaster for a TV station in Cookeville informed us that the rain had become mixed with snow there. Cookeville is located on the Eastern Highland Rim (Fig. 2) at an elevation of 1133 feet. A few other reports of sleet also were received, further indicating that the freezing level was lowering.

Shortly after the Cookeville call, a spotter reported it was snowing in Gainesboro with a visibility of 1/2 mile. Gainesboro is also located on the Eastern Highland Rim (Fig. 2). Shortly thereafter, rain changed to snow at Crossville with a visibility of 1/2 mile. Crossville is located on the Cumberland Plateau (Fig. 2) at an elevation of 1882 feet. It was becoming very obvious that the forecasts were in BIG TROUBLE and needed to be amended, and soon.

A call was placed to NWSFO Memphis to relay the information, and allow them some time to diagnose the situation and issue updated forecasts. During the discussion, the author mentioned that dynamic cooling may be responsible for the changeover occurring so soon, similar to what happened during the December, 1997 snowstorm mentioned above. The forecaster at Memphis agreed that dynamic cooling might be an explanation for the changeover, especially when considering that surface temperatures to the north over central Kentucky were slightly warmer, and dew points were only a few degrees lower than those observed over middle Tennessee. This would rule out the possibility that colder and/or drier air had advected into middle Tennessee, thus causing the changeover to occur earlier than forecasters expected.

The 0000 UTC February 4 sounding for Nashville (not shown) depicted a freezing level of about 2200 feet MSL, or 1600 feet AGL. Using this knowledge and information from two studies (AWS 1979) that related the probability of snow at the surface to the height of the freezing level AGL, the picture became clearer why the changeover was occurring so soon. The table below derives from those two studies.

>1200 feet
Most precipitation
will be rain
920 feet 50%
660 feet 70%
315 feet 90%

At the time the sounding was taken, it was not yet snowing at NWSO Nashville (elevation 591 feet). To the east however, because of the elevation increase on the Eastern Highland Rim (Cookeville's elevation is 1133 feet), given a similar low level temperature structure, the freezing level there was probably 1000-1100 feet AGL. Apparently this was low enough for the changeover to begin. Further to the east at Crossville on the Cumberland Plateau (elevation 1882 feet), the freezing level was probably 300 to 400 feet AGL, giving a high probability of snow reaching the surface.

During the early evening, NWSFO Memphis issued a Winter Storm Warning (4 to 8 inches of snow) for mainly the northern Cumberland Plateau counties, and a Snow Advisory (1 to 3 inches) for counties immediately to the west through south of the warning area.

Figure 3 shows a composite radar view of 0.5 base reflectivity, including the estimated precipitation type valid 2345 UTC February 3, 1998. This figure was obtained from WSI Corporation via the Internet. In a discussion with a WSI employee, it was learned that the normal 0.5 base reflectivity imagery is enhanced using a computer program that uses surface observations, surface temperatures, dew points and wet bulb temperatures to differentiate between rain, mixed rain and snow, and snow. Sometimes the precipitation type may not be correctly differentiated but an image such as this may cause a meteorologist to investigate what is going on, especially if only rain is in the forecast.

Using the NGM as an example, Figs. 4 through 11 compare the 24-hour forecasts valid 1200 UTC February 4, 1998, with the NGM initial analyses for that same time (two model runs later). This comparison identifies the gross underforecasting of the strength of the storm system as well as the additional cooling that resulted because of the increased upward vertical motion. Following is a summary.



Figs. 4 & 5 - 1) Sea Level Pressure of storm 988 mb or 6 mb lower than forecast.

2) 1000-500 mb thickness about 30 meters lower than forecast over middle Tennessee.

Figs. 6 & 7 - 1) 850 mb low 60 meters lower than forecast.

Figs. 8 & 9 - 1) 700 mb low 90 meters lower than forecast.

Figs. 10 & 11 -1) 500 mb low 60 meters lower than forecast.

Figure 12 shows the MOS guidance for Crossville based on the 1200 UTC February 3 NGM run. During the event, the conditional precipitation type forecast (PTYPE) was rain, and the conditional probability of snow (POSN) was low. These two parameters clearly showed the poor performance of the NGM MOS during this event. Finally, Fig. 13 shows the reported snowfall amounts for the approximate 12-hour period ending around 1230 UTC February 4. The highest snowfall amounts occurred on the northern Cumberland Plateau, where 12 inches or more of snowfall was reported from several counties. Since the low-level winds were generally from the north during the event, and the Cumberland Plateau is oriented from NNE to SSW, it appears orographic lifting was not a factor in the increased snowfall totals on the plateau.

Using the AVN model run for 1200 UTC February 3, 1998 (not shown), which had the more accurate forecast of the intensity of the storm, a few of the reasons for the higher snowfall totals on the Cumberland Plateau include the following:

5. Conclusion

This paper briefly discussed a winter storm that caused much hardship for the people of eastern middle Tennessee. It appears dynamic cooling played a major role in changing rain to snow far sooner than anyone expected. This type of event is difficult to anticipate. Sometimes only a degree or two of cooling may be all that is needed to change rain to snow and produce a major winter storm. However, by keeping in mind the three suggestions that follow, it may be possible to at least recognize the potential for dynamic cooling, and then word forecasts accordingly if the precipitation type may be affected.


The author wishes to thank Mike Murphy, service hydrologist at NWSO Nashville, for comments and suggestions to improve this paper; also Scott Sharp and Shannon White, meteorologist and meteorologist intern respectively at NWSO Nashville, for obtaining the snowfall information used in Figure 13.


AWS , 1979: Meteorological Techniques, AWS Pamphlet 105-56, Headquarters AWS,

Scott AFB, IL, 2-3.5-1.

Geddes, A. E. M., 1939: Meteorology. 2nd Ed. Blackie & Son Limited, 160.

Steigerwaldt, H., 1986: Deformation zones and heavy precipitation. NOAA Technical

Memorandum NWS CR-83, NWS Central Region, SSD, Kansas City, MO.