February 12, 2001 Winter Weather Event and
the "Seeder - Feeder Mechanism"
Background information on the February 12, 2001 Winter Weather Event.
On February 12, an ice storm, consisting primarily of sleet, was expected over the
RAH CWFA. The heaviest sleet was expected during the morning between 12Z and 18Z.
In retrospect, the precipitation was considerably lighter and more of a mixed variety than
expected due to poor model performance prior to the event. Easily lost in the larger scale,
however, was a mesoscale snow outbreak, which occurred overnight prior to the main
precipitation event. Upon examination, it was determined that the snow resulted from a
very pronounced seeder-feeder mechanism in which low level supercooled clouds were
seeded with ice crystals from above. This seeding modified the mid and low level
temperature profiles, glaciated the low level clouds, and produced light snow for several
hours in the foothills and northwest piedmont. Snowfall amounts were not large in RAH's
CWFA, ranging from ½ to 1 inch in the northwest piedmont, and were overshadowed by
the larger scale icing event which evolved shortly thereafter. It is clear, however, that the
seeder-feeder mechanism alone could easily produce an "advisory-scale" event,
consisting of either frozen or freezing precipitation. The purpose of this paper is to explain
how the seeder-feeder mechanism might be foreseen by providing a real-life and close-to-home example for future reference.
To review very briefly, the 'seeder-feeder' mechanism is the introduction of ice from
above into a lower level liquid or supercooled liquid cloud. This introduction of ice
provides condensation nuclei, thus initiating precipitation from the low level cloud layer.
The low cloud layer may consist of 1) liquid droplets, 2) supercooled liquid droplets, or 3)
if the temperature is cold enough (maximum low layer temperature of < 1C - figure 1) the
cloud may glaciate, and produce snow. The resulting precipitation type is, of course,
dependent upon the thermal profile from the cloud to the surface (as well as temperatures
of exposed surfaces in the case of freezing rain).
The icing event is not the focus of this paper, thus we will not provide an analysis
of synoptic scale features leading up to the event. The seeder-feeder mechanism, being
mesoscale in size, would likely not be readily discernable beyond 24 hours. We will thus
narrow our analysis to a time frame beginning 18 hours prior to the onset of precipitation
at 12Z on the 11th, and concentrate mainly on forecast soundings and observed data.
Initial Mesoscale Conditions at 00Z - 2/12/01
From the attached zone forecast excerpt (figure 2), we were expecting the
precipitation (sleet) to begin towards morning as late as the evening update on Sunday
night, February 11. At sunset, western NC was blanketed by a uniform deck of
stratocumulus based from 5 to 6 thousand feet, while an altocumulus deck around 10
thousand feet was approaching from the southwest. (figure 3). The 00Z Greensboro
sounding (figure 4) was quite dry, and a significant portion of the sounding was warmer
than 0C below 10 thousand feet. Of particular note is a warm (> 0) 2 to 3 thousand foot
thick layer at the mid-levels (790-710 mb, ~7 to 10 thousand feet). The 00Z radar
composite (figure 5) showed a narrow band of light rain over western SC, where Anderson
was the only METAR site reporting rain. The rain was evidently in response to a small
vorticity max, depicted on the 00Z Mesoeta over the Greenville-Spartanburg area, which
was forecast to move quickly northeast across western NC. Temperatures at 04Z across
the area were in the 40 to 45 degree range (figure 6) and, given significant precipitation
amounts, would fall rapidly due to diabatic cooling as dewpoints were in the teens. This
would produce wet bulb temperatures in the 30 to 35 degree range.
What Occurred Overnight
The light rain moved into the southern piedmont near Charlotte shortly before
midnight. This initial surge of precipitation was not expected to be significant, as evidenced
by the lack of its inclusion in the evening update. The precipitation's early arrival would not
appear to pose a problem for the forecast of sleet, as diabatic cooling would quickly lower
surface temperatures to freezing or below as the rain moved north of Charlotte into the
foothills and northwest piedmont. Meanwhile, warm air advection, as evidenced by the
veering wind profile at GSO (figure 4) and temperature profiles from downstream sites at
Nashville and Atlanta (figure 7), would reinforce and even strengthen the warm air already
in place at the mid-levels. We could then expect a changeover to primarily sleet as
forecast. Catching us somewhat by surprise, the precipitation began to fall as snow (figure
8). Observations at Hickory and Statesville showed the precipitation beginning at 05Z as
rain, then changing to snow by 06Z and remaining snow until around 09Z. At Greensboro,
the precipitation began as snow at 08Z and continued until around 1030Z when it changed
over to freezing rain, which lasted through the afternoon. The snow accumulated up to an
inch in the northwest piedmont and foothills.
Diagnosing the 'Seeder-Feeder' Mechanism Responsible for the Snow
The GSO sounding at 00Z shown in figure 4 points towards a sleet/freezing rain
scenario, as it exhibits an 80mb deep warm layer aloft (790 - 710 mb) as well as the
potential for significant diabatic cooling at the surface. If we go beyond a cursory glance,
however, there are a couple of subtle details to note as well: 1) The warm layer, while
deep, is only 1 to 2 degrees above freezing and very dry - thus we should expect diabatic
cooling here as well; 2) we see the initial hint of mid-level moisture at 560 mb, just above
the -10C isotherm (snow formation likely); 3) the maximum warm layer temperature (note -
the warm layer is the lower cloud layer, not the entire boundary layer) is around -2C at
Upstream soundings at Nashville and Atlanta in figure 7 show that the mid- level
moisture which will be arriving in the CWFA has a base around 750 mb (~ 8 kft) and is
extremely deep. Notice also that the -10C isotherm is around 550 mb, so little thermal
advection will occur at this level downstream at GSO.
06Z GSO Sounding
At the 06Z sounding (figure 9), the mid-level moisture has deepened and now has
a base below 600 mb, with the vast majority of the layer colder than -10C. The mid-level
melting layer has cooled to less than 0C as well, and only the surface layer remains
At this point, the low level cloud has glaciated and snow began to fall in GSO
shortly before 07Z. METAR observations of ceilings indicate that the mid-level altocumulus
deck had progressed to GSO by 06Z. A few hours prior, rain had begun in the Charlotte
area around 04Z, shortly after the arrival of the mid-level cloudiness. It is apparent that the
mid-level cloudiness was responsible for seeding the clouds below, thus causing the
unexpected surge of precipitation prior to the principal, more widespread precipitation.
Approximately one hour separated the arrival of the mid-level cloud deck and the onset of
snow at GSO.
Comparison of Observed Soundings and ETA Forecast Soundings
Forecast soundings from the ETA model's 12Z run on February 11 and 00Z run on
February 12 were analyzed and compared to the actual soundings that were taken during
the event. The presence of a warm nose, any above freezing layers, low level winds, and
dry layers were noted.
The 12Z model sounding initialized well with a warm nose between 850 and 700mb
and little moisture in the sounding (not shown). At 18Z (figure 10), the model sounding
correctly showed increasing moisture, especially above 400mb, and continued to show a
deep above-freezing layer. The model sounding did not show the stratocumulus deck at
Figure 11 compares the ETA 12 hour forecast sounding valid at 00Z with the
observed sounding at GSO. The observed sounding showed that the upper level moisture
(above 400mb) continued to increase, with a thin moist layer evident at 560mb. This layer
was the leading edge of the mid-level cloud deck that was approaching from the
southwest. The model sounding underestimated the moisture above 400mb, and did not
pick up on the approach of mid-level cloud deck from the southwest at 560mb. The
observed sounding showed the presence of a stratocumulus deck at 800mb, while the
model had a 15 degree dewpoint depression at this level. A dry 2°C warm nose remained
between 800 and 700mb in the forecast sounding. This compared well to the actual 00Z
sounding, although the model greatly underestimated the amount and depth of the dry air
from 800 to 600mb.
These model inaccuracies had large implications on its p-type forecast. The
observed cloud layer that was approaching from the southwest near 560mb, which was
not depicted in the model sounding, began to precipitate into the dry warm nose between
00Z and 06Z. This caused evaporative cooling in that layer, and reduced the temperature
of the warm nose to near freezing isothermal. Because the model underestimated the
amount of dry air that the precipitation was falling into (between 800 and 600mb), the
model underestimated the potential for evaporative cooling in that layer. The precipitation
from the mid-level clouds proceeded to seed the stratocumulus deck at 810mb with ice,
which later produced snow. Since the model did not account for the stratocumulus deck,
the mid-level cloud layer, or the very dry air between 800 and 600mb, it did not correctly
portray the seeder-feeder process.
The 6 hour forecast sounding valid at 06Z is shown in figure 12. At this time, the
model did pick up on the stratocumulus deck near 810mb and the precipitating mid-level cloud layer.
At 06Z, the actual sounding (figure 9) showed an isothermal layer at
0°C between 750 and 700mb with the rest of the sounding below freezing, except at the
surface. However, the model showed that the layer was not near-freezing isothermal; it
remained near 2°C. Because the model forecast the layer to be too warm
(underestimated the amount of evaporative cooling), the model p-type forecast of sleet
and freezing rain at 06Z was incorrect. A warm nose between 1°C to 3°C warm nose
will produce a snow/sleet mix if ice is introduced, and freezing rain if ice is not
introduced (see figure 1, which was taken from the VISIT training session "P-Type
Forecasting - The Top-Down Approach"). Since the ETA did not account for the mid-level cloud layer that introduced ice into the stratocumulus deck between 00Z and 06Z,
the model did not account for ice seeding into the lower cloud layer early enough, and
freezing rain was the predominant p-type that was forecast by the model.
Evaluation of partial thickness scheme and TRENDS
The 1000-850mb thicknesses and 850-700mb thicknesses were
computed from the actual soundings and compared to the ETA model forecast
thicknesses. The ETA forecast thicknesses were taken from the model run closest to
that time. The partial thicknesses from the observed soundings and the forecast
soundings were plotted on a partial thickness nomogram. The observed thickness
nomogram correctly portrayed snow changing to mainly freezing rain and a little sleet.
However, the model forecast thicknesses painted a different p-type scenario - one that
was dominated by only freezing rain and sleet. Very slight differences in the observed
and forecast soundings led to the inaccurate model forecasts.
The following chart compares the actual partial thicknesses and their
corresponding p-type based on the nomogram with the model forecast thicknesses and
||Observed p-type from
||ETA forecast p-type from
||Mostly FZRA, trace PL
||Mostly FZRA, trace PL|
||Mostly FZRA, trace PL
||Mostly FZRA, trace PL|
||Snow if isothermal near
||Mostly FZRA, trace PL|
||Mostly FZRA, trace PL
||Measurable PL w/ FZRA|
Figure 13 shows the low level thicknesses from the observed soundings plotted
on a nomogram. The corresponding ETA model thicknesses were plotted on a separate
nomogram, which is shown in figure14. Note that at 18Z, the ETA overestimated the
thickness of the above-freezing layer and underestimated the low level thickness.
Although the model correctly lowered the thickness in the above-freezing layer at the
initialization time of 00Z, it was still too strong compared to reality. From 00Z to 06Z,
the observed 850-700mb thickness decreased slightly, from 1553m to 1549m, placing
the sounding in the "snowy nose" category of the nomogram. In this area of the
nomogram, the p-type will be snow if a near-freezing isothermal layer is present;
otherwise freezing rain and sleet will be the dominant p-type. In the observed sounding
at 06Z, a near-freezing isothermal layer was present, and snow was able to reach the
ground. At the same time, the model forecast 850-700mb thickness did not change,
while the 1000-850mb thickness dropped by 11m. A near-freezing isothermal layer was
not present in the 06Z forecast sounding. Instead, the ETA forecast sounding showed a
2°warm nose between 800 and 700mb. This overestimation of the above-freezing layer
led to an incorrect model p-type forecast of mostly freezing rain with a trace of sleet.
The model overestimated the amount of low-level cold advection that was occurring
and overestimated the temperature of the above-freezing layer, leading to a freezing
rain and sleet forecast.
Between 06Z and 12Z, the model forecast 850-700mb thickness dropped 5m, while
the observed drop in 850-700mb thickness was between 00Z and 06Z. The model may
have been too slow to forecast evaporative cooling in this layer, which was due to the
absence of the mid-level clouds that approached GSO at 00Z and precipitated into the
above- freezing layer.
The partial thickness nomogram performed very well in predicting p-type, even
though the thickness changes were very subtle. While the ETA forecast thicknesses were
close to the observed thicknesses, they were off just enough to cause an incorrect p-type
forecast. The model underestimated the effects of evaporative cooling in the warm nose
layer because it did not account for the presence of the mid-level clouds, which
precipitated into the layer and seeded the stratocumulus deck.
The feeder-seeder mechanism can be diagnosed by carefully examining the entire
depth of the soundings. Very small changes in the vertical temperature profile can
profoundly affect the p-type, and models often cannot resolve these details. Therefore,
observed data must be analyzed intensively to improve the forecast. Comparing the
differences between the observed sounding and the forecast sounding is critically
important to determine when the model is deficient or accurate. The presence of any cloud
layers or dry layers that the model is not accounting for must be carefully considered when
making a p-type forecast.
This sort of event is not particularly difficult to diagnose beforehand - IF one is
familiar with the mechanism responsible and thoroughly analyzes the observed and
forecast soundings accordingly. This case study was drafted for such a purpose - to show
that the seeder-feeder mechanism does exist and can cause potentially significant
precipitation type problems, and that it can be forecast, at least in the short term.