Map depicting area of
interest on May 13, 2004. Shaded counties are Milam and Robertson.
Following the passage of a strong cold front more than 10 days prior to the event, return flow from the Gulf commenced on the 4th and continued without interruption through the morning of the 13th. Moisture return was gradual, but by the evening of May 12th dew points were ~22° C across the area and had reached ~ 25° C along the immediate coast, ~95 miles to the south. Surface analyses from NCEP did not depict, nor did objective and manual analysis of surface observations reveal any surface boundaries near the area.
On the preceding eveing, objective analysis of (00 UTC) upper air data revealed a mid and upper level synoptic scale trough west of the area over the Rockies. A weak mid-level trough was advancing eastward ahead of the primary trough. Streamline and isotach analysis revealed a jet stream maximum exceeding 70 knots over the Great Plains and a subtropical jet maximum exceeding 70 knots over northern Mexico. Between the two jets, a broadly diffluent southwesterly flow covered much of Texas.
The area that became the center of the excessive rainfall event on the 13th lies in the southeastern corner of the County Warning Area (CWA) of the NWS office located in Fort Worth. The afternoon zone forecast package (ZFP) on the 12th placed the probability of measurable precipitation at 60% for the daylight hours on the 13th. The area forecast discussion (AFD) mentioned that warm advection would develop over most of the CWA by midday Thursday and heavy rainfall would be possible with the approach of a cold front, mainly from late Thursday into early Friday (the 14th). The evening AFD update, (issued around 8pm CDT) opined that no precipitation was expected “overnight” . Indeed, the operational numerical models were not at all alarming with respect to QPF output, at least through noon on the 13th, as neither the ETA nor the GFS produced precipitation that exceeded flash flood guidance across the area. The GFS model (from the evening run) did produce a large area of very intense precipitation after 1pm CDT on the 13th, but over far eastern Texas.
In this study we
utilized a variety of platforms, some in situ and some remote, to document
the development and evolution of this event. These included automated surface
observations, products from the WSR-88D radars, wind profiles from profilers
operated by NOAA and by the Texas Commission on Environmental Quality,
rawinsonde observations, and satellite imagery from the eastern GOES unit.
We also made use of the 0-hour (analysis) output from the Rapid Update
Cycle (RUC) model as a source of pseudo-real-time gridded data above the
surface. The figure below shows the area of the study, including county
lines, radar and profiler sites, and selected surface observation sites.
Map depicting area of
interest on May 13, 2004. Shaded counties are Milam and Robertson.
The RUC 0-hour (analysis) output for 4am (09 UTC) revealed a continuation of the deep southeasterly flow of warm, moist air into the area. Perhaps of greater significance, divergence at 250 mb had increased and the isotach analysis at that level depicted a zone of stagnating flow over the area between Austin and Houston. The figure (below the IR satellite image) shows a zone of less than 30 knot flow oriented southeast to northwest from approximately Galveston to Waco. The most intense and rapidly expanding deep convection was within this stagnation zone.
Just before 5pm
CDT (10 UTC), the Hydrometeorological Prediction Center (HPC) updated
the Excessive Rainfall Discussion (QPFERD) to outline an area over
southeastern Texas into Louisiana where isolated excessive rainfall
would be possible. The southwestern corner of the area was Hearne (LHB),
in Robertson County. The issuance mentioned that latest model guidance
pointed to a 12 to 16 hour period of isolated heavy rainfall across the
region, and noted that the latest satellite and radar observations were
“very supportive”, with increasing convection through south central Texas
that was directly related to subtropical shortwave energy lifting northeastward.
The 3-hour accumulated precipitation product from the WSR-88D at KGRK at 7:15am (1215 UTC) (not shown) indicated a large area of rainfall exceeding 2 to 3 inches northeast of the radar site, with a small area near and west of Hearne showing more than 4 inches. Very heavy rainfall was continuing across the area, as seen in a composite of 0.5° base reflectivity from KGRK at 7:03am CDT (1203 UTC) with the 7am (12 UTC) surface observations plotted in the usual manner. Base velocity products (0.5 degree) from KGRK between 11 UTC and 13 UTC (not shown) indicated the development of very strong east to west outflow from the area of intense rainfall over eastern Milam and western Robertson counties toward the KGRK radar site. Around 7am (12 UTC) a significant area of greater than 50 knot inbound velocity was shown from near Cameron to near the KGRK site. In loops of the velocity product, this strong outflow can be seen passing across the GRK site, with the flow only gradually subsiding. Then, by 8am (13 UTC), strong convergence developed very near the KGRK site associated with yet another (but more limited) burst of deep convection.
Infrared imagery
from the eastern GOES satellite depicted the signature of a developing
mesoscale convective complex in the same area, with an expanding area of
very cold cloud tops over Milam, Robertson and southern Falls counties,
as seen in the figure immediately below the composite radar plot. The RUC
0-hour (analysis) output for 7am (12 UTC) revealed a continuation of the
deep southeasterly flow of warm, moist air into the area, with expansion
of the previously noted area of divergence at 250 mb (figure immediately
below the IR satellite image). The isotach analysis at that level depicted
a zone directly over the area of deep convection where flow had decelerated
to less than 20 knots.
Both the one-hour and the three-hour accumulated precipitation products from KGRK are missing or unusable (from the NCDC archive) for the period from approximately 8am (13 UTC) to almost 11am (16 UTC). However, review of reflectivity data from KGRK and other WSR-88D sites leaves no doubt that very heavy rainfall persisted during that entire period.
Infrared imagery
from the eastern GOES satellite at 10:15am (1515 UTC) (second figure below)
depicted a mature MCC, with a very large area (>110,000 km2)
in which cloud tops were at or below –52° deg C. An interesting feature
seen in animation of the IR imagery between 6am (11 UTC) and 12 noon (17
UTC) was a cyclonic looping of the coldest cloud tops over and west of
the zone of most intense rainfall. The RUC 0-hour (analysis) output for
10am (15 UTC) (third figure below) revealed a continued expansion of the
area of divergence at 250 mb. The isotach analysis depicted a zone directly
over the area of deep convection where flow was less than 20 knots, and
depicted the area of maximum divergence to be centered west of Lufkin.
Infrared imagery
from the eastern GOES satellite at 1:15pm (1815 UTC) (second figure below)
depicted some weakening of the MCC, with significant shrinking of the area
of coldest cloud tops, and a reorganization of the coldest tops into a
more linear formation east and southeast of Bryan-College Station into
a more north to south orientation. The RUC 0-hour (analysis) output for
1pm (18 UTC) (third figure below) relocated the most intense divergence
at 250 mb westward from the 10am (15 UTC) position to near Centerville
(in Leon County) and showed further intensification of the divergence.
The isotach analysis indicated that the area of stagnation (where flow
was less than 20 knots) was shifting southeastward away from central Texas
toward the Texas coast near Houston.
A variety of surface
data was examined in an attempt to generate ground-truth measurements of
accumulated rainfall. Probably the most reliable reading was obtained from
the Hearne Airport (KLHB) AWOS. Inspection of the archived on-line
observations transmitted by this station revealed a number of data outages
during several periods between 12 UTC and 16 UTC. Fortunately, the unit
does maintain a 30-day archive on-site. The unit includes a tipping bucket
precipitation measuring system which resets to zero after each 20-minute
automated report. With the assistance of the manufacturer (Vaisala), the
complete data record for May 13th was recovered, and it revealed a 12-hour
rainfall at Hearne of 12.21 inches (for the period ending at 2005 UTC).
The table below the figure displays the 20-minute, 1-hour, 3-hour, and
total precipitation accumulation for the period 0805 UTC through 2005 UTC.
.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table showing 12-hour precipitation
at Hearne, TX (KLHB) on May 13, 2004. Accumulation is shown in 20-minute
increments for running one-hour and three-hour accumulations, and total
accumulated precipitation. Red numerals indicate maximum accumulations
for 20-min., 1-hour and 3-hour periods.
Fortuitously, the Ledbetter (LDBT2) site was located only ~45 miles south of the heaviest rainfall in this event, and thus in a position to provide important information regarding the evolution of the event on the inflow side. The figure below shows the 60-minute resolution wind data from LDBT2 covering a 14-hour period from 07 UTC to 21 UTC on May 13th. The 11 UTC data is missing from the graphic. The profiler archive for the period was dumped to a text file which permitted recovery of the 11 UTC data. The 0-3 km data is summarized in the table below the figure.
Perhaps the most
striking feature of both the figure and the table is the significant acceleration
in southerly flow that developed around 11 UTC. Between 10 UTC and 11 UTC,
the lowest gate (621 m) shows the wind backing from 175 degrees to
157 degrees while the velocity increases from 34 to 43 knots. By 12 UTC,
the first gate is measuring 48 knots, with >40 knots indicated at all gates
through 3 km. This intense southerly flow continues until gradually abating
around 15 UTC.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table showing 60-minute average wind
data from Ledbetter, TX (LDBT2) NOAA profiler.
(Note: time in UTC; height in kilometers
above MSL; wind speed in knots).
By ~12 UTC, the flow at KGRK is significantly different from that seen at LDBT2, with a southeasterly flow of 20 knots at the lowest gate, turning at the next gate to south-southwest at 30 knots and gradually veering to southwesterly through 3 km. The highest velocity is ~30 knots. By ~13 UTC, the lowest gate at KGRK shows the arrival of the strong (~25 knots) east-northeasterly flow associated with the westward-migrating rain-cooled outflow mentioned previously in Sec. 4. The flow turned to the southwest at the remaining gates through 3 km, at a velocity no greater than 35 knots. Thereafter, a gradual decrease in the 0-3 km flow was indicated.
The KEWX WSR-88D
VAD wind profile data was inspected and it reflected that the strongest
low-level flow occurred shortly after 12 UTC but was <35 knots. Around
2 km, the flow was briefly ~40 knots. However, the data did not reflect
the intense, deep southerly flow seen at LDBT2. Data from CLETX depicted
a more south-southwesterly flow that did not reach 40 knots at any time.
Data from the NOAA profiler at Palestine (PATT2) was missing for the period
08 –15 UTC.
IR satellite imagery for this convective system was analyzed according to the foregoing specifications. Maximum extent of the cold cloud tops (satisfying both of the requisite specifications) was observed at 1645 UTC, when cloud top temperatures of <-33°C covered an area of approximately 3.25 X 105 km2 within which was found a region where cloud top temperatures were <-52°C covering an area of 2.12 X 105 km2. This system was, therefore, slightly more eccentric than the Maddox specification.
Fritsch and Forbes (2001) also define the life cycle phases of MCCs, relying in part upon earlier work by Maddox (1980) and Zipser (1982). The stages, in sequence, are as follows: initiation, development, mature, and dissipation. The initiation stage covers the time from formation of the first storm(s) until the IR cold cloud tops first satisfy the minimum required for an MCC; the development phase begins at that point and continues until the maximum extent of IR cold cloud tops (as specified previously) is reached; the mature phase begins at that point and continues until the extent of IR cold cloud tops no longer satisfies the minimum required for MCC definition; the dissipation phase begins at that point and continues until convection ends.
In this case, initiation occurred ~0730 UTC just east of Austin (AUS) as storms fired just behind the southwest flank of a precursor MCS that was centered between College Station (CLL) and Houston (HOU). The development stage began at ~1115 UTC when IR satellite imagery indicated that the minimum requisites for an MCC (temperatures and extent) had been met. The mature stage began at ~1645 UTC, as noted above.
Fritsch and Forbes (2001) identified certain thermodynamic patterns and dynamical features that are usually present when MCCs develop. These include pronounced low-level convergence in the terminus region of developing low-level jets, where vertical motion associated with the low-level convergence contributes significantly to destabilization of the local environment prior to the onset of deep convection. Lifting and destabilization are especially strong when a low-jet intersects the thermally-direct circulation associated with frontogenetic forcing. In addition, a weak mid-level short wave may be approaching the genesis region and such a system further enhances low-level convergence associated with the low-level jet. Finally, warm advection usually dominates the lower troposphere while diffluent flow is found in the mid and upper troposphere.
Examining this event in light of the foregoing, we find that while there was a modest low-level jet in place, the event did not occur near its terminus. As noted earlier, analyses through 06 UTC revealed no sign of a low-level baroclinic boundary in the region where the MCC developed. Similar analyses at 09 UTC were also negative for such a boundary. It is possible that the first MCS, which moved across the area from Austin to Houston just prior to the initiation of convection that became the MCC, may have created a diffuse boundary that drifted northward within the broad south-southeasterly flow, helping to precondition the lower troposphere for additional deep convection. There was a weak mid-level short wave advancing toward the genesis region ahead of the synoptic scale trough which was still well to the west. And as noted earlier, warm advection did dominate the lower troposphere with a mildly diffluent flow aloft.
Fritsch and Forbes (2001) note the apparent importance of the low-level jet, normally a nocturnal feature, in the development of MCCs. Citing a number of sources, they posit that “… nocturnal low-level jets develop as a result of adjustments that take place as 1) the mixed layer decouples from the surface as the surface cools and 2) horizontal temperature differences develop as a result of sloping terrain (e.g. the Great Plains) and an east-west gradient in the Bowen ratio (the ratio of surface sensible to latent heat flux).” They note that these processes are independent of the dynamics of migratory disturbances, so the low-level wind accelerations produced by the nocturnal low-level jet provide a significant enhancement to the low-level warm advection and convergence that would normally be present as a result of an approaching short wave and/or synoptic scale circulation.
This event did not occur on the Great Plains, and although it initiated during a nocturnal period, it persisted well into the middle of the following day. With respect to the importance of sloping terrain and its involvement in the generation of nocturnal low-level jet features, the topographic realm of central and southern Texas is significant. The higher terrain of the Rocky Mountains retreats westward across western Texas, but then reappears west and southwest of Del Rio in the form of the Sierra del Burro Mountains, and continues southeastward through Coahuila and the western portions of Nuevo Leon. These terrain features (part of the Sierra Madre Oriental), although not rising as high above sea level as the mountains of Colorado and northern New Mexico, generate a significant gradient eastward toward the coastal plains of central and southern Texas, as well as eastern Nuevo Leon and Tamaulipas.
Fritsch and Forbes (2001) suggest that MCC events can be classified into one of two types. The type-1 events involve slantwise ascent above a surface-based front or baroclinic zone. Type-2 events occur in warm sector environments without the presence of synoptic scale frontal forcing, relying instead upon downdraft-generated cold pools originating from deep convective storms typically rooted in a well-mixed boundary layer. As downdraft-generated cold pools expand, they supply layer lifting and a source for slantwise ascent. Convective overturning in type-2 events can be downshear, upshear or remain upright, depending upon the relative strengths of low-level vertical wind shear and the downdraft-generated cold pool. Unlike the type-1 events, in type-2 events the slantwise front-to-rear ascent does not begin until after a cold pool develops. Our analysis of this event points to a type-2 classification, primarily due to the absence of a low-level baroclinic boundary. In addition, the evolution of the event suggests that downdraft-generated cold pools were critical to the evolution of the MCC.
Fritsch and Forbes (2001), citing Maddox et al. (1986), Houze et al. (1990), and Tollerud and Collander (1993), note that if severe weather occurs in association with an MCC, it usually occurs during the initiation stage. There were several warnings for severe weather during the event (both for possible tornadoes and for damaging straight line winds), and most of the documented severe weather events occurred during the transition from the initiation stage to the development stage. They also cite McAnelly and Cotton (1989) and Collander (1993) for the proposition that the heaviest rainfall typically occurs in the development stage, which was certainly the case in this event. They further note that the heaviest rainfall is usually concentrated on the equatorial flank of the coldest cloud shield, which is often the southwestern flank. Examination of the WSR-88D products and the IR satellite imagery in this event is consistent with that pattern.
A number of authors (see Fritsch and Forbes, 2001) have proposed that long-lived mid-level mesovortices (also called mesoscale convective vortices, or MCVs) with warm cores are thought to be an inherent process characteristic of MCCs. The dynamical structure and circulation of these MCVs are said to resemble those of tropical disturbances. A close examination of animations of both the GOES-12 IR imagery and of the 0.5° base reflectivity products from KGRK reveals the evolution of features associated with at least one MCV with this system.
The KGRK 1203 UTC
0.5 degree base reflectivity product is suggestive of a spiraling series
of banded convective elements arranged around a central point. There
is, however, an area of little or no activity in the vicinity of College
Station. About the same time, the GOES-12 IR product shows an
interesting arrangement of cloud-top temperatures, with
an area of warmer cloud tops near College Station (CLL). These two
features are seen both before and after this time, but were more prominent
prior to this point in time. The authors propose the possibility
that this feature represents
an “inflow notch” for the MCV, where
very strong inflow and ascent were displacing convective elements northward
(in a fashion reminiscent of inflow notches sometimes seen with supercell
thunderstorms).
Another interesting feature, seen in both products but more apparent in the animation of the radar product, is the appearance that the most intense convection (coldest cloud tops in IR) make a cyclonic loop over several hours time between ~11 UTC and ~15 UTC. The authors have no current explanation for this apparent evolution, but it is hypothesized that more than one MCV may have developed with this MCC, with the possible interaction of two MCVs to produce the looping of the most intense convection as seen by radar and of the coldest cloud tops. We intend to follow up on the latter hypotheses in subsequent research, perhaps making use of the 0-hour RUC analyses in an attempt to locate and track the mid-level vortices.
Finally, Fritsch and Forbes (2001) discuss the ability of MCCs to modify the local, regional and (occasionally) the synoptic environment. Recall that this event took place in advance of a major synoptic-scale trough that was advancing eastward into the southern plains. Significant severe weather events were expected on May 13th as the dynamics of the advancing trough interacted with the warm, moist and unstable low-level environment over Texas, Oklahoma, Missouri and Kansas. The MCC appears to have directly interfered with the expected evolution of severe weather on the 13th.
Severe weather events
were actually focused (see figure below) in an arc extending from Wichita
Falls to Abilene to Junction to San Antonio, which generally represents
the westward limits of the cold outflow generated by the morning MCC over
central Texas. The Texas events primarily involved marginally severe hail,
while the northeastern Oklahoma events were predominantly strong straight-line
wind events. Very few severe events were reported in Kansas and Missouri.
We propose that the intense MCC that occurred exhausted the low-level instability
over a large area of Texas via a variety of processes, including convective
overturning, intense cold pool development and spreading outflow, and disruption
and obstruction of the warm, moist and unstable southerly flow from of
the Gulf of Mexico. As a result, the extent of the severe weather associated
with the primary upper trough was greatly diminished.
Fritsch, J.M., and G.S. Forbes, 2001: Mesoscale convective systems. Severe Convective Storms, Meteor. Monogr., Amer. Meteor. Soc., 323-357.
Houze, R.A., B.F. Small, and P. Dodge, 1990: Mesoscale organization of springtime thunderstorms is Oklahoma. Mon. Wea. Rev., 118, 613-654.
Maddox, R.A., 1980: Mesoscale convective complexes. Bull. Amer. Meteor. Soc., 61, 1374-1387.
__________, K.W. Howard, D.L. Bartels, and D. M. Rodgers, 1986: Mesoscale Meteorology and Forecasting, P. Ray, Ed., Amer. Meteor. Soc., 390-413.
McAnelly, R.L., and W.R. Cotton, 1989: The precipitation life cycle of mesoscale convective complexes over the central United States. Mon. Wea. Rev., 177, 784-808.
Tolerud, E.I., and R.S. Collander, 1993: A ten-year summary of severe weather in mesoscale convective complexes. Part I: High wind, tornadoes, and hail. Preprints, 17th Conf. on Severe Local Storms, St. Louis, Mo., Amer. Meteor. Soc., 533-537.
Zipser, E.J., 1982: Use of a conceptual
model of the life cycle of mesoscale convective systems to improve very-short-range
forecasts. Nowcasting, K. Browning, Ed., Academic Press, 191-204.