These training sessions were developed in the pre GOES-R era and contain dated information. Although some of the principles may still be applied in the GOES-R era, these are no longer supported training courses and are made available here as reference material. Be sure to check the link to “Training Sessions” for current, supported courses.
The following narrative summarizes the initial formation period of Tropical Storm Katrina which intensified to a Category 1 hurricane at landfall near Miami, Florida, and became a Category 5 hurricane in the Gulf of Mexico. This is an excerpt from the Tropical Prediction Center’s Tropical Cyclone Report by Knabb et al (2005)
The complex genesis of Katrina involved the interaction of a tropical wave, the middle tropospheric remnants of Tropical Depression Ten, and an upper tropospheric trough. This trough, located over the western Atlantic and the Bahamas, produced strong westerly shear across Tropical Depression Ten, causing it to degenerate on 14 August approximately 825 n mi east of Barbados. The low-level circulation gradually weakened while continuing westward, and it eventually dissipated on 21 August in the vicinity of Cuba. Meanwhile, a middle tropospheric circulation originating from Tropical Depression Ten lagged behind and passed north of the Leeward Islands on 18-19 August. A tropical wave moved through the Leeward Islands and merged with the middle tropospheric remnants of Tropical Depression Ten on 19 August, forming a large area of showers and thunderstorms north of Puerto Rico. This activity continued to move slowly northwestward, passing north of Hispaniola and then consolidating just east of the Turks and Caicos during the afternoon of 22 August. Dvorak satellite classifications from the Tropical Analysis and Forecast Branch (TAFB) of the Tropical Prediction Center (TPC) began at 1800 UTC that day. The upper tropospheric trough weakened as it moved westward toward Florida, and the shear relaxed enough to allow the system to develop into a tropical depression by 1800 UTC 23 August over the southeastern Bahamas about 175 n mi southeast of Nassau. The depression was designated Tropical Depression Twelve rather than “Ten” because a separate tropical wave appeared to be partially responsible for the cyclogenesis, and, more importantly, the low-level circulation of Tropical Depression Ten was clearly not involved.
The two GOES satellite animations illustrate the evolution of Katrina in the formative stages during the four day period of August 21-24, 2005. The water vapor images show the flow associated with the upper trough. The upper vorticity center within the trough is clearly seen, as it tracks along to the west to the north of the pre-Katrina system. The second satellite animation includes visible daytime images along with the 3.9 micrometer nighttime images. The tropical wave and the mid-level remnants of TD10 can be tracked as it undergoes cyclogenesis and becomes Tropical Storm Katrina.
The images suggest that the active deep convection that produced the cold IR areal maximum near 21N 74W at 0815 UTC 23 August played an important role in Katrina’s genesis. The low-level circulation center that became Katrina’s center either originated or abruptly intensified due to this convection. Fig.1 shows the location of this convective maximum on the 0815 UTC image along with the location of the upper level vorticity center. The “best track” analysis (Knabb et al 2005) has the first Tropical Depression designation at 23.1N 75.1W at 1800 UTC 23 August, and first named storm location at 24.5N 76.5W at 1200 UTC 24 August. Fig. 1 shows those locations along with the track of the pre-existing disturbance (dashed) and the best track positions through the end of the animation periods at 2345 UTC 24 August.
During the late morning of 22 May 2008, thunderstorms formed north and east of the Denver Metro area and moved northward into northeast Colorado. Storm motion was actually west of due north, meaning storms forming in the climatologically favored region near Denver International Airport tracked toward the Urban Corridor of Colorado’s Front Range (along and west of I-25), rather than a more typical eastward motion. One of the first storms of the day turned out to be the most intense, tracked over Windsor, Colorado, and produced a relatively intense tornado and large hail. The grand majority of tornaodes forming so close the Front Range are non-supercell tornadoes (or “landspouts”), and typically do little to no damage. This tornado has been rated EF-2 or EF-3 based on a damage survey completed by the National Weather Service Boulder office.
The two figures below show the 500 mb and surface analyses at 1800 UTC (noon local time) on 22 May. An intense closed 500 mb low was centered near Las Vegas, and a few notable jet streaks rotated around it. One such jet streak was located in Eastern Colorado, placing the northern Front Range in the favored left-exit region. At the surface, an extremely deep low was centered near Denver with relatively dry air south and east of it, and a tongue of moist air was being advected westward north of a roughly east/west-oriented boundary.
The GOES-12 Visible Loop (click the image below for the animation) clearly shows the boundary between dry air to the south and moist air to the north. As the moist air noses to the southwest toward the north side of Denver between 1400 – 1600 UTC, the stratus clouds just behind the boundary dissipate, allowing a narrow region to receive full sunlight. Clicking on the second image below shows an annotated visible image from 1615 UTC with the location of this boundary denoted with a black line. The storm of interest intiates just south of this boundary around 1645 UTC, and really blows up around 1700 UTC as it reaches the warm, moist air.
Mesoscale vortices across the Great Lakes during the winter are observed as arctic air is warmed from below by the relatively warm water. Forbes and Merritt (1984) discuss lake-effect vortices over the western Great Lakes, where they are more commonly observed relative to the eastern Great Lakes. Laird (1999) investigated the existence of coexisting mesoscale lake-effect vortices over the western Great Lakes. This discussion will focus on a rare occurence over the eastern Great Lakes that occured on December 13, 2005.
The synoptic scale conditions favorable for wintertime mesoscale lake-effect vortices include 1) large difference between the lake and low-level air temperature, 2) convergence over the lake, 3) unstable environment at low-levels and 4) weak surface flow and horizontal pressure gradients (Forbes and Merritt 1984; Hjemfelt 1990). These conditions existed on December 13, 2005. Great Lakes water temperatures were in the 4 to 10 degrees Celsius range over the eastern Great Lakes. The 12:00 UTC surface analysis shows a 1028 mb high just northeast of the Great Lakes leading to relatively weak surface winds and horizontal pressure gradients. Also, note the convergence over the lakes caused by the land breeze circulation.
An inspection of the 12:00 UTC sounding from Alpena, MI shows an unstable environment at low-levels. The difference between the 850 mb and lake surface temperature was approximately 22 degrees Celsius. This is much greater than 13 degrees which is considered the minimum to initiate lake-effect snow (Holroyd, 1971) and corresponds to a dry adiabatic lapse rate.
The GOES-12 visible satellite loop shows simultaneous mesoscale lake-effect vortices over Lakes Huron, Erie and Ontario. Over Lakes Erie and Ontario, convergence boundaries caused by the land breeze circulation develop vortices. Over Lake Huron multiple convergence boundaries exist.
The 10.7 um IR imagery views this event from earlier on (06:15 UTC) through the daytime. The vortex over Lake Huron is readily observed at night. We also observe the convergence boundaries form over the south shoreline of Lakes Erie and Ontario then advect northward before developing vortices. The MODIS visible image from 18:10 UTC provides us with a higher resolution (0.5 km) compared to the 1 km visible GOES imagery shown above. This is a true color image which allows us to see additional details such as sediments and ice cover over the lakes.
In the early afternoon of 24 May 2005, thunderstorms formed in northeast Colorado ahead of a weak shortwave trough embedded in moderate westerly upper-level flow. Surface dew points were mostly in the 50’s F, with slighly higher values near the eastern border. A number of storms showed supercellular characteristics, but were embedded within a larger convective mass. Severe storm reports included high wind, a number of large hail reports, and several tornadoes.
Satellite observations of these storms include GOES-East and GOES-West, in addition to a MODIS (Aqua) pass around 2010 UTC. This allows the comparison of thunderstorm top structure between GOES and MODIS. GOES-12 was in Rapid Scan Operation (RSO), so approximately 8 images per hour are available, but the relatively coarse spatial resolution prevents analysis of the fine-scale structures atop the storm’s anvil. MODIS has 36 channels with a 1-km footprint, and several of the shorter wavelength bands have even finer resolution (down to 250-m).
The GOES-12 visible loop shows the explosive development of thunderstorms around 1900 UTC. Morning convection in extreme northeast Colorado left behind several boundaries which are most evident around 1930 UTC; these boundaries no doubt played very important roles in the evolution of the convection, and perhaps in tornadogenesis. The GOES-12 10.7um loop also shows evidence of the boundaries, as well as a rapid cloud-top cooling rate, indicating a very unstable atmosphere.
Landfalling tropical cyclones pose a number of serious threats, including high wind, storm surge, flooding rainfall, and tornadoes. During the unusually active tropical season in 2004, Hurricane Ivan produced over 100 tornadoes in the US resulting in 8 deaths. Evacuations in the vicinity of the landfall location no doubt saved many lives, but tornadoes often occur hundreds of kilometers from the landfall area, and forecasting their exact location hours ahead of time is impossible. As a result, forecasters must issue tornado warnings in the traditional “nowcasting” timeframe, i.e., minutes ahead of time.
Tornadoes associated with landfalling tropical cyclones typically occur in very high shear, low instability situations, and therefore their radar characteristics differ greatly from tornado-producing classic supercells. Relatively fewer large-ice particles result in lower reflectivity values, and the horizontal extent of the tornado-producing cells is often small. A small hook-shaped appendage can sometimes be found with these cells, and their rotation can occasionally be detected by the radar’s radial velocity output. Nonetheless, the subtle radar signatures with these tornadic cells often result in a large false-alarm-ratio, so having additional data sources should aid in the warning decision-making process.
On Wednesday, September 15, Hurricane Ivan was approaching the Gulf coast of Florida and Alabama, and rainbands in its northeast quadrant produced 29 tornadoes which caused all 8 deaths. The GOES-12 Infrared Loop on 15 September shows Ivan’s approach, and convective bursts in the Florida panhandle and southwest Georgia are evident based on rapidly cooling cloud tops. As will be shown below, almost every tornado occurs with one of these bursts of convection.
In the absence of high-level cirrus clouds, visible imagery can be used to identify more precisely the locations of each of these convective bursts. The 1-km-resolution GOES-12 Visible Loop shows many overshooting tops emerging from the lower deck of clouds. They are especially evident in the late afternoon because their associated shadows become more pronounced with the low sun angle. These overshooting tops can be more easily identified by viewing the visible loop (above). It should be noted that GOES-12 was in “Rapid Scan Operation” (RSO), meaning 8 images per hour, compared to the usual 4, were available over the eastern US. Without this improved temporal resolution, these individual overshooting tops would be more difficult to identify.
In order to examine a possible relationship between convective bursts and tornado reports, we will look more closely at several different times. The first three tornado reports occurred around 2045 UTC: one in Gulf County, FL, and two in Early County, GA. The 2040 UTC Visible Image shows the approximate location of the tornadoes (at the tips of the red pointing arrows) and the associated overshooting tops. There is a slight northward shift of the overshooting top relative to the tornado report location; this is due to parallax. The map on these images is drawn relative to the ground. Storm tops over 10,000 ft. MSL in the midlatitudes result in noticable parallax and the map should actually be shifted northward around 7 km so the tornado locations correctly fall underneath the overshooting tops.
The 2040 UTC Infrared Image shows the cloud top associated with the Gulf Co., FL, storm has a brightness temperature of -69 C (the coldest in the area), and between 2025-2040 UTC, this cloud top cooled from -57 C to -69 C. The brightness temperatures with the Early Co., GA, storm are warmer, but they represent a local minimum in the immediate area. It should be mentioned that the Gulf Co. storm moved northwestward and an associated tornado struck Panama City Beach, resulting in damage to a commercial area and 1 fatality.
Next, we will look at two tornado reports from Calhoun county in GA from 2120-2130 UTC. The 2132 UTC visible image shows the approximate location of a tornado report in southwestern Calhoun County at 2130 UTC. Notice the overshooting top evident just to the north of the report; again, parallax correction places the overshooting top directly over the location of the tornado. The 2132 UTC infrared imageshows a cloud top with a brightness temperature of -51 C, another local minimum in the area. This cloud top continued to cool with time (see the IR loop above), reaching a minimum at 2155 UTC.
At 2125 UTC, another tornado was reported in Panama City, FL. However, in this case, neither the visible nor the infrared imagery showed a well-defined overshooting top or cold cloud top. The radar data (below) shows convection in the area but the precise location of rotation is difficult to pinpoint.
The final tornado we will examine occurred at 0223 UTC near Blountstown in Calhoun Co., Florida. Four fatalities were reported, making this the single deadliest tornado of the day. Since this occurred well after dark, only the 0225 UTC infrared image will be shown. A convective burst was observed around 0100 UTC southwest of Tallahassee near the coast; cloud tops continued to cool and this convective complex moved (interestingly) northeastward. The tornado report occurred underneath -70 C cloud tops at 0223 UTC, shortly before the cloud tops began to warm.
Radar loops from KEVX (Eglin AFB, FL panhandle) and KTLH (Tallahassee, FL) are provided below. Both base reflectivity and radial velocity are shown. In the KEVX reflectivity loop, step forward to 2039 UTC (approximately the same time of the initial Gulf Co. tornado report). Between 2039-2104 UTC, five distinct cells, each with hook-like signatures, can be seen approaching the Florida coast (Bay, Walton, and Okaloosa counties). Notice also that velocity couplets can be found with several of these cells. However, only the southeastern most of these cells was associated with a tornado report. Recall from the visible loop that only this southeasternmost cell had a distinct overshooting top. This single example suggests that looking for satellite features such as overshooting tops can improve the warning false-alarm ratio. Finally, it is quite possible that these other cells were also producing waterspouts and/or tornadoes, but they may not have been reported; a large number of residents had already evacuated from the this portion of the Florida panhandle.
On May 4, 2003, a thunderstorm formed along a dry line in southeastern Oklahoma. It quickly split, and the left-mover traversed across Oklahoma before interacting with other supercells in northeastern Oklahoma and Missouri. Click on the GOES-12 visible loop below. Some interesting features of this storm include:
As a side-note, it’s always interesting to compare the location of jet maxima to the water vapor satellite imagery. Below is the 18:15 UTC GOES-12 water vapor image with the model analyzed 18 UTC 300mb wind speed (left) and 500mb wind speed (right). Notice that the 300 mb jet max lies further north, which is expected because upper level troughs are tilted toward colder air. Also note that the 300 mb jet max is north of the gradient in water vapor brightness temperatures, but the 500 mb jet max is aligned along this gradient. Similar observations should be made in the future to determine if this is a consistent trend.
Below, you will find a more detailed analysis of the left-moving thunderstorm. I plan to write this up more formally and submit it to Weather and Forecasting. These are some of the initial ideas.
On May 4, 2003, severe thunderstorms erupted across eastern Oklahoma, eastern Kansas, Missouri, Arkansas, and Tennessee. There were 94 tornado reports, in addition to hundreds of large hail and severe wind reports. This discussion will focus on one thunderstorm which formed in southeast Oklahoma and produced severe hail but no tornadoes.
The 18Z sounding from Oklahoma City shows a steep lapse rate above an inversion at 800 mb. The boundary layer is quite moist, and very dry air is located above 800 mb. Surface observations at 22Z indicate dew points in the low 70’s and southerly winds east of Tulsa (KTUL). The wind profile in the sounding shows south-southwesterly low-level flow, veering to west-southwesterly at mid- and upper-levels. Both the low level and deep layer shear profiles are conducive for severe thunderstorm development. Wind profiler data from Haskell, Oklahoma, (see the map below left) which is very near where the storm of interest passed, shows midlevel winds increasing through the morning hours and low-level winds from the south. The 0-3 km hodograph from the 21Z profiler data is slightly curved in a clockwise direction in the lowest 1 km, but is nearly straight in the lowest 3km.
The GOES-12 water vapor loop shows a negatively-tilted upper-level trough across Kansas and Oklahoma, with dry mid-level air moving into eastern Oklahoma. The GOES-12 visible loop shows towering cumulus forming along a dryline in eastern Kansas by 19:10 UTC and in eastern Oklahoma by 20:15 UTC. By 20:45 UTC, a mature storm can be seen in southeastern Oklahoma, and by 21:15 UTC, this storm has split. The left-mover from this split is the storm of interest.
Note the location of the left-mover at 21:32 UTC, then note its location 1 hour later at 22:32 UTC. During this one-hour period, the storm progressed 81 miles, so it averaged 71 knots during this portion of its lifetime. Compare this to right-moving storms, like the storm which forms in northeastern Oklahoma and moves almost due east. It’s moving at 42 knots, slightly more than half the speed of the left-mover. Additionally, the left-mover’s motion vector is 220 degrees, compared to 250 degrees from the right-movers. Surface winds are from the south, but since the left-moving storm’s motion has a southward component, its storm-relative inflow is approximately 50 knots from the north. This strong inflow of moist, buoyant air will likely lead to an intense updraft due to enhanced convergence along its leading edge.
There is quite a bit of literature addressing the deviant motions and intensities of splitting supercells. Of recent interest, Bunkers and Klimowski (2000) discuss why left-moving storms in an environment where the shear profile lies in the upper-right quadrant of the hodograph are expected to move faster than their right-moving counterparts. When the hodograph is curved in a clockwise direction, the right-mover is expected to have a stronger updraft due to pressure perturbation forces. However, in a very unstable environment, left-movers can be quite strong and produce severe weather. In this case, the hodograph is almost straight, and the left-moving storm produced 2.75 inch diameter hail and wind gusts greater than 50 knots. Severe winds are no surprise, considering the storm and the leading edge of its cold pool were moving at greater than 70 knots. The gust front can be seen in the visible loop preceeding the storm and extending westward from its leading edge.
It should be mentioned that the storm’s motion can also be affected by its own outflow (Bunkers and Klimowski, 2000). Recall the dry air in the 800-500 mb layer seen on the 18 UTC sounding. Precipitation falling into this dry air would produce a large amount of evaporation, which would drive a strong downdraft and an intense gust front. Its motion, therefore, is likely determined by the mean wind, internal storm dynamics, and the nature of its cold pool.
Another interesting feature about the visible loop above is the orientation of the anvil. Notice at 22:10 UTC how the anvil of the left-mover is oriented in a west-northwest to east-southeast direction, while the anvils of the right-movers are oriented southwest to northeast. This seemingly odd feature should actually be anticipated: anvil motion depends not only on the anvil-level winds, but also on the motion of the storm itself. [An analogy is a steam locomotive traveling down the tracks when the wind is blowing orthogonal to its motion – it will leave behind “puffs” of steam which are advected by the wind from the point where they were released.] Anvil motion can be calculated by taking the vector difference of the storm’s motion from the anvil-level wind vector. Doing this simple geometric calculation using the left-mover’s motion vector and anvil level winds of 100 knots at 255 degrees predicts anvil motion at 299 degrees, which is almost exactly what is observed. This observation suggests that left-movers can be pinpointed with satellite using only the orientation of the anvil at a snapshot in time.
This discussion is only meant to introduce this interesting left-moving storm which occurred on a day where much more significant severe weather was happening elsewhere. More in-depth analysis is certainly possible. In particular, the left-mover continues to the northeast and interacts with at least 2 other right-movers. This interaction is very complex and probably quite important. I plan to investigate this phase of the storm’s life before submitting the journal article.
Thanks to Jim Purdom, John Weaver, Jack Dostalak, and Louie Grasso for ideas and insight.
Bunkers, Matthew J. and B. A. Klimowski, 2000: Predicting supercell motion using a new hodograph technique. Wea. Forecasting, 15, 61-79.
On 29 April 2003, a stationary front extended across east-central Colorado, with southerly winds and relatively dry air south of the front, and moist easterly winds north of the front (see the 1800 UTC surface map below). By mid-afternoon, dew point values near 50 (deg F) were in place north of the boundary in eastern Colorado. Due to the expectation of severe weather, GOES-10 rapid scan operation (RSO) was called.
The GOES-10 Visible Loop shows a northwest/southeast oriented cloud boundary around 1840 UTC along the northern edge of the Palmer Lake Divide (a higher terrain feature which extends eastward into the Colorado plains – see the image above). Around 2000 UTC, towering cumulus clouds were evident along this boundary, and by 2030 UTC a storm anvil is visible. Shortly thereafter, the storm appears to split. As the right-mover intensifies, notice the cumulus clouds feeding in from the southeast. By 2200 UTC, the storm has taken on definite supercellular characteristics: a super-crisp edge on the south and west sides of the anvil, vertical cloud wall on the west side of the storm (extremely bright due to reflected sunlight), and a nice flanking line extending from beneath the southern portion of the anvil. The storm continues to progress eastward (to the right of the wind shear vector), and eventually dies around 0030 UTC. Note that before it begins to dissipate, around 2245 UTC an arc-shaped line of cumulus clouds is moving southwestward into Eastern Colorado from Kansas. This boundary separated unstable air to the west from more stable air to the east. The stable air mass originated from a cloudy region in northwest Kansas. The storm dissipates shortly after intersecting this boundary and ingesting the cooler, more stable air.
The KFTG radar loop (Denver’s NEXRAD radar) shows storm initiation near Limon, Colorado, which is about 55 miles from KFTG (for Limon’s location, see the first map above – Limon is LIC). It quickly split, and the right mover became the dominant storm (verifying what we observed on satellite). At 2200 UTC, the visible satellite indicated an impressive-looking supercell; the radar signature is more marginal, with maximum reflectivities near 45 dBZ. At this time, the storm is 60 miles from KFTG, so at 0.5 degree tilt, we’re looking at 4700 ft AGL. In other words, the radar does not give us a good idea how intense this storm actually is.
The second radar loop is from the Goodland, Kansas, NEXRAD (KGLD), approximately 90 miles east of the storm’s location at 2230 UTC. By 2245 UTC, the radar signature is much more impressive, with a 60 dBZ core and a nice inflow notch. We can also see the southward moving boundary noted in satellite which intersects the storm around 0015 UTC, shortly before the storm’s rapid dissipation.
The two photos below were shot near Seibert, Colorado, around 2315 UTC. This is a perfect example of an LP supercell, with a bell-shaped updraft base and relatively little precipitation visible. The storm also produced 2.25 inch diameter hail during this portion of its life.
This example shows a case in which radar signatures were not impressive early in the storm’s life, but visible satellite imagery indicated an intense supercell. Severe thunderstorm warnings accompanied the storm throughout most of its lifecycle, and a few tornado warnings were issued due to strong mid-level rotation evident on radar (not shown).
On April 16, 2003, the passage of a sharp cold front (temperature dropped 56 deg F at Blue Hill Observatory in less than 18 h) was associated with the development of an undular bore off the Massachusetts coast, as noted by the GOES 12 visible satellite loop below:
Long, thin bands of low cloud form parallel to the front (probably indicated by the leading cloud arc) and travel southward across the Cape Cod region. East of the Cape, the pattern is better defined with 7 or 8 narrowly spaced bands. In this case, the wave train was excited when the cold front interacted with a strong marine layer inversion caused by the advection of unseasonably warm air out over the chilly Atlantic. Record highs of 84 and 85 were set at Boston and Blue Hill Observatory, respectively, while Chatham out on Cape Cod (just below the northward bend of the Cape), remained in the 50’s. This marine inversion is shown nicely in the skew-T plotted from Chatham’s rawindsonde observation:
For comparison, the postfrontal skew-T at 00Z on the 17th is also shown above, about 4 hours after the frontal passage.
Following are some surface observations from Chatham during the time window of interest (the most recent obs are at the top):
From the satellite images, it appears that the leading cloud arc passes through Chatham at 2015Z, which coincides nicely with a windshift observed at the surface at 2017Z and a peak wind of 31 kts at 2020Z. Pressure was rising rapidly during the cold front passage.
This images from this event show some similarities to the undular bore observed on the Texas coast a few years ago, as chronicled in a ‘Picture of the Month‘ feature in Monthly Weather Review (Clarke, 1998).
In both cases, it appears that a sharp cold front interacted with a strong temperature inversion ahead of the front to cause the bore. In the Texas case, it was a nocturnal inversion. In the Massachusetts case, it was a strong marine boundary layer inversion. In the Texas case, the cloud system was not associated with surface temperature or dewpoint change, just a prefrontal pressure trough and wind, whereas the bore in the Massachusetts case appears to be nearly coincident with the front.
For more on undular bores, see Smith (1988).
Clarke, J. Christopher 1998: An Atmospheric Undular Bore along the Texas Coast. Mon. Wea. Rev., 126, 1098-1100.
Smith, R. K. 1988: Traveling waves and bores in the lower atmosphere: The “Morning Glory” and related phenomena. Earth Sci. Rev., 25, 267-290.
This case study is being presented to demonstrate what might be learned about the pre-tornadic storm environment in and around Colorado and Kansas on 31 May 1996. While closely linked with the earlier case in this series on this outbreak, for this presentation we are limited to the more standard mode of GOES satellite operations, e.g., imagery at 15 minute intervals. Both GOES-9 and GOES-8 digital, McIDAS formatted data, that cover the area and time of interest, can be found on the CIRA-RAMM Team’s FTP server. Log on to “canopus.cira.colostate.edu” (or “220.127.116.11”), using “anonymous” and then your e-mail address for the password. The data is located in the “/96152_1min/prestorm/” directory.
The GOES-9 datasets include 26 visible images from 11:00:13 to 18:00:13 UTC (AREA1040 – AREA1065), 58 IR images from 03:00:13 to 18:00:13 UTC (AREA3908 – AREA3965), 58 water vapor images from 03:00:13 to 18:00:13 UTC (AREA4008 – AREA4065) and 37 “fog product” images (e.g., difference images from the 10.7 and the 6.7 micron channels) from 03:00:13 to 12:00:14 UTC (AREA6908 – AREA6944).
GOES-8 datasets are available as follows: 16 visible images from 17:01:xx to 21:15:xx UTC (AREA8861 – AREA8876) and 16 IR images over the same time span (AREA4861- AREA4876).
Morning analyses on 31 May 1996 found a weak longwave trough over the western U.S. and a number of significant low-level boundaries in the central Plains. These included a warm front in eastern Kansas and Oklahoma; an east-west oriented outflow boundary in Kansas, analyzed as part of the warm front by NCEP (NWS’ National Center for Environmental Prediction) on the MSLP (mean sea-level pressure) analysis; a surface low in southeast Colorado; and a convergence boundary along a trough line which stretched from the low into southeastern New Mexico. The airmass over most of the central plains was unstable and expected to become even more so as the day progressed, due to diurnal heating and moisture advection from the Gulf of Mexico.
Depending upon your needs, please review the “basic”, first level, or the “advanced”, more detailed, discussion of the precursor environment to the tornadic outbreak. Better yet, visit both of them!
Upper-level tropospheric disturbances are characterized by alternate regions of upward and downward motion. Because the atmosphere warms (cools) as it subsides (ascends), the 6.7 micron imagery channel (which is sensitive to the temperature of upper level water vapor) can be used to identify and track such features. Study the animated 6.7 micron imagery and try to locate shortwave signatures approaching the area of interest. Larger-scale features such as jet streams are also indicated by this imagery. They often appear as sharp gradients in brightness temperatures on a large scale. Can you find the subtropical jet?
Which, if any, lower-tropospheric features can be seen using the 6.7 m imagery? This channel is sensitive to water vapor at mid- to upper-levels, as well as to cold upper-level cloudiness. Thus, we should not expect to see lower-level atmospheric phenomenon. Careful perusal of sequential imagery from this period shows features that seem to be associated with the high terrain of the Rocky Mountains. Can you explain this apparent contradiction? For information on this, as well as the other 4 imaging channels, refer to the CIRA-RAMM Team’s Introduction to the GOES-8 Imager tutorial.
Use sequential, 10.7 micron imagery to study the squall line which occurs overnight in Nebraska and South Dakota. Obviously, rain-stabilized air (i.e., low-level outflow) is being left behind by these storms, but is not observable over most of the region due to overlying anvil cirrus. However, the area at the south end of the line is not anvil-covered. (Is an outflow boundary being left behind in this region?) If it is, why can’t we see it?
Now observe the same sequence using the “fog product” imagery (found in AREAs 6908-6944). The fog product is made by differencing the 3.9 and the 10.7 micron channels, and is sensitive to water clouds (show as light gray in the enhancement used here). Notice the line of lighter gray clouds to the south and west of the squall line. These are stratiform clouds forming along and above the outflow airmass, and can be seen quite clearly in this product. The boundary is not as easy to see on the 10.7 micron channel, since the actual temperature of the cold ground and stratiform clouds are similar. (10.7 micron is a window channel which senses the actual tempertures of viewed scenes.) What is influencing the extent/coverage of this cloud mass? Is terrain influencing the shape of this cloud area? For an in-depth discussion of this useful product, refer to the CIRA-RAMM Team’s GOES’ 3.9 micron Channel tutorial.
The outflow boundary, once identified, can be followed into the morning hours on visible imagery. Track the location of this feature for as long as you can. HINT: look for differences in cloud type on either side of the feature. Which channels, if any, can be used to differentiate cloud type? Much of the stratiform cloudiness dissipates shortly after sunrise. What happens to the cloud fields near this boundary after the low clouds dissipate? What clues can you see in the imagery to confirm whether or not the outflow boundary is still present?
Mean sea-level pressure analyses from NCEP continued to analyze a generally north-south oriented warm front in eastern Oklahoma and Kansas throughout the day. This front may be seen on the introduction page to this section. While a distinct boundary could be followed using surface observations, no clearly identifiable cloud line can be seen on GOES imagery. Why is this? What ramifications does this factor have for later convection? HINT: Initial insolation may be evaporating surface moisture from overnight precipitation. Is this likely in these areas from what can be seen in the satellite imagery? Once a cloudy region clears off, what differences would you expect to find in skin temperature values in 10.7 micron imagery between it and regions which were not cloudy? Will air temperatures reflect these differences immediately?
The 6.7 micron channel on the GOES imager is sensitive to water vapor in the middle and upper levels of the troposphere. Frequently, atmospheric disturbances (such as shortwaves or jet streaks) can be tracked using this channel, because the zones of upward and downward motion associated with these features remain quasi-steady in a storm-relative sense. Note, for example, the comma-shaped feature moving across southern Colorado. This is a shortwave trough that played a role in aiding developing convection. Notice how much easier it is to see this feature on the 6.7 micron channel compared with the 10.7 micron channel.
Lower tropospheric boundaries are important to the development and evolution of deep convection. One of the more important boundaries on this day was an east-west oriented outflow boundary situated in central Kansas at the time of convective development. Look at both the 10.7 micron and fog product imagery (AREA’s 6908-6944) from overnight. This derived-product is made by differencing the 3.9 micron and 10.7 micron channels, because the brightness temperature at 10.7 micron is greater than that at 3.9 micron for water clouds. Notice the line of stratiform cloudiness to the south and west of the squall line traveling through Nebraska and South Dakota. These clouds are along the outflow boundary being produced by the storms, and can be easily identified using the fog product. The boundary is not nearly as easy to see on the 10.7 micron channel, since the actual temperature of the cold ground and stratiform clouds are similar.
The outflow boundary, once identified, can be followed into the daylight hours. Construct an animated sequence of visible images to follow this feature, and see how late you can track it. HINT: look for differences in cloud type on either side of the feature.
For those with RAMSDIS units, compute cloud motions for the smaller cumulus clouds in central Kansas. These cloud motions reveal the location of a moderately strong convergence zone that may play a role in later convective development. Stratiform clouds in southeast Colorado can also be used to track the circulation around the low in the early part of the morning. Be careful, though, because dissipation later in the period can make cloud-derived motions deceiving.
Interesting weather in the region on the 31st of May includes: 1) a line of tornadic thunderstorms that formed along a dissipating warm front in northern Kansas; 2) tornadoes also in eastern Colorado near the point where a dryline, a cold front and the warm front intersected; and 3) large hail reported in Kansas, Colorado and Texas. For more details about what the imagery can tell us about this meteorological event, please visit one, or both, of our highlight summary pages depending upon whether your interest is basic or advanced, technically.
This GOES-8 dataset from 31 May 1996 contains 265 files, consisting of one file for each of the 5 imaging channels, at full resolution, for each of 53 separate scanning periods. The observations occurred between 21:15:15 and 22:59:18 UTC. As can be seen from the inventory, the elapsed time between scanning sequences varies, but during the periods 22:04:12 – 22:11:49 and 22:35:12 – 22:42:49, 30-second interval sequences are available. The 22:04:12 VIS and the 22:35:12 VIS images can be seen above, left and right. They represent the first of the 15 images available in each of these two 30-second sequences.
The digital, McIDAS format data are available by FTP, using either “18.104.22.168” or “canopus.cira.colostate.edu”. Log in with “anonymous” and then provide your e-mail address as the password; the data are in the 96152_1min/earlydev directory.
When inspecting the cloud fields using the 10.7 micrometer (µm) imagery, notice how the lower clouds are warm, while the higher clouds and thunderstorm tops are cold. Also notice that the low-level cumulus, which are small and bright in the visible (VIS) imagery, and warm in the 10.7 µm infrared (IR), can be seen in all imaging channels except the 6.7 µm IR (mid-level water vapor-sensitive) channel. This is because the signal at 6.7 µm comes from water vapor that is located in the middle and upper layers of the atmosphere, above the height of the low-level cumulus.
The normal imaging frequency for GOES is every 15 minutes. This special research data set contains periods when imagery was taken at 15-min., 1-min. and 30-sec. intervals. Create animated loops from the image sequences at the different time intervals and note the differences in your ability to follow cloud features and to observe thunderstorm development. Notice how the cumulus clouds can be followed much easier with the more frequent-interval imagery, due to their relatively short life-span. Notice too how small-scale features at the thunderstorm tops, such as waves and overshooting tops, can be observed more easily with more frequent-interval imagery.
The thunderstorms in the central Plains exhibit a noticeable degree of organization and have formed in preferred (expected) locations. The main activity is associated with an east-west aligned warm front in central KS, a low pressure area in eastern CO and a low-level convergence zone (often called a dry-line) which stretches from western KS into western TX. Use both single images and loops to study these boundaries and the storm development along them.
It is instructive to observe the thunderstorms using all the imager’s channels and at very frequent intervals. Note how the overshooting tops in the VIS imagery correspond with colder areas in the anvil at 10.7 and 6.7 µm. These regions represent the location of active updraft areas of the thunderstorms and will be associated with regions of radar echo. See how the anvils of the thunderstorms extend downwind from the cold, overshooting area and that, above some of the anvils, there are long plumes of cloud. These long cloud plumes seem to emanate from the updraft area and, when they are long and continuous, indicate a long-lived thunderstorm as their generator. Longer-lived storms are often severe, when other atmospheric conditions allow.
Notice how the clouds appear differently in the 3.9 µm channel imagery. Higher, cirrus clouds are dark, while lower, cumulus clouds are bright. This is a complex subject which is covered in some detail in the CIRA-RAMM Team’s GOES 3.9 µm Channel Tutorial.
Track the motion of the different thunderstorms using the animated imagery. See how the large storm, which develops in central CO, moves more to the south than the other storms in the region? Such anomalous storm motion (often termed a “right mover”) may indicate the existence of a storm with severe weather associated with it; in this case, a tornado.
Look at the low-level cumulus fields in detail, using all imaging channels. Identify the differences (e.g., why can’t the low-level cumulus be seen in the 6.7 µm imagery?). Can GOES imagery (all five channels) be utilized to identify and position different air masses, as well as the convergence boundaries separating them? Can you locate a dryline using the multispectral characteristics of the imagery? Is it colocated with the surface position of the dryline; and, if not, why?
This data set contains data collected at a much more frequent interval than the normal operating mode of every 15 minutes. Construct loops at the different imaging frequencies and note how much easier it is to track the clouds and to understand and anticipate storm development and evolution with the higher frequency images. What would be the optimum frequency interval for several different applications?
Once thunderstorms have developed, use the higher frequency imagery along with WSR-88D data to look at them and their surrounding environment. Do these two data sources, used together, enhance your understanding of thunderstorm intensity as compared to using either one alone?
Observe the cloud top temperatures in the 10.7 µm IR window channel. Compare this with available radar data and determine what portion of the storm contains an active precipitation core. Compare the coldest regions at anvil top, as well as other portions of the anvil, with imagery from the other IR channels and look for differences. If there are, can you explain why they exist, and their meaning, with respect to active portions of a thunderstorm?
In your comparisons of IR and radar imagery of cloud tops, as suggested above, did you take into account the satellite parallax viewing and radar beam size and height-above-surface at the different viewing distances? Try using 30-sec. satellite imagery with 5-min. radar data.
Note that the developing thunderstorms in the central Plains form in preferred locations. The main activity is associated with an east-west warm front in central KS, a low pressure area in eastern CO and a dryline which stretches from western KS into west TX. Use individual frames and loops to study these boundaries. Specifically, look for clues of imminent storm development (e.g., VIS loops can help identify boundary development and the location of the center of the low).
Track the motion of storms using satellite imagery loops and notice the rightward deviation of the large storm in central CO. Use available supplementary data to compute how the storm-relative helicity for this storm differs from others in the region.
Use sequential imagery to track the cirrus clouds and compare your results with profiler data, if available. Identify cumulus and mid-level clouds and do the same thing. Finally, compare the cumulus motions with the motions that can be determined from the WSR-88D data.
Notice how some of the cumulus change from bright to dark in the 3.9 µm imagery. Do you know why? (You may wish to refresh yourself on the nuances of this unique imaging channel by reviewing the CIRA-RAMM Team’s GOES 3.9 µm Channel Tutorial.) Are there corresponding differences in the WSR-88D imagery when bright-to-dark changes in the 3.9 µm channel imagery occur? What about electrical activity reported by lightning detection systems?
The LES event covered by this imagery affected both the upper and lower portions of Michigan. This case was chosen because of the availability of super-rapid scan operation (SRSO), 1-minute interval, imagery collected during the morning hours when both single- and cross-lake snow bands were occurring. One-hour snow accumulation totals included 2-3″ at Wakefield, in the far northwestern part of the state; 2-3″ between 1600-1700 UTC at Bellaire, in the northwestern portion of the lower part of the state; 2-3″ at Montague, in the western portion of the state’s lower part; and 2-3″ between 1700-1800 UTC at Traverse City, in the northwestern portion of the lower part of Michigan.
Sample six-hour accumulations (from 1200-1800 UTC) from the eastern parts of upper Michigan were > 8″ at Shingleton; > 5″ at Seney, and 2″ at Melstrand. Other nearby sites, including SSM, HTL, MKG, GRR, LAN, and FNT, reported trace amounts only, for the entire 6-hour period.
For more information on the nature of lake-effect snow, please refer to a report on a major 63″ LES event.
The digital, McIDAS format data files are available by anonymous FTP to: “canopus.cira.colostate.edu”. They are in directory 97016_LES. The 5-channel, full one-minute sector files are named with the conventional McIDAS “AREAnnnn” format, with nnnn=1000’s indicating GOES Imager channel 1, nnnn=2000’s for channel 2, etc. Channel 1 is visible, channel 2 is centered at 3.9 micrometers, channel 3 is at 6.7 micrometers (water vapor), channel 4 is the conventional IR window, and channel 5 is at 12.5 micrometers. Sets of the individual “AREAnnnn” files are also available in “zipped” format, collected in 143 self-extracting *.EXE files, with each of these containing all five channels for a 1-minute time period. These zipped, “LES_????.EXE” files, can be accessed directly by clicking here if your browser’s “helper” option for .EXE files, under the Options/General – Preferences menu, is set to “SAVE”.
The LES datasets cover a 23-hour timespan, from 0345 UTC on 16 Jan. through 0245 UTC on the 17th (with the VIS available from 1145-2255 UTC on the 16th). The SRSO period begins at 1104 UTC, with AREA?042 / LES_009.EXE, and ends at 1711 UTC, with AREA?153 / LES_102.EXE.
FASTEX took place during January and February of 1997. It focused on low pressure systems over the North Atlantic region and information from aircraft flights, surface observations, dropsondes, radiosondes, radars and satellites was collected and archived. Meteorologists from many different European and North American countries participated.
The CIRA-RAMM Team coordinated with scientists in Europe in requesting special one-minute frequency satellite scanning operations from NOAA’s GOES-8. From these special datasets a subset, covering a low pressure system moving across the Atlantic Ocean during the period 1245 – 1845 UTC on February 5, 1997, is now available for access on CIRA’s Virtual Laboratory server.
These digital, McIDAS format data files are available by anonymous FTP to: “canopus.cira.colostate.edu”. They are in directory 97036_FASTEX. The 5-channel, full one-minute sector files are named with the conventional McIDAS “AREAnnnn” format, with nnnn=1000’s indicating GOES Imager channel 1, nnnn=2000’s for channel 2, etc. Channel 1 is visible, channel 2 is centered at 3.9 micrometers, channel 3 is at 6.7 micrometers (water vapor), channel 4 is the conventional IR window, and channel 5 is at 12.5 micrometers. Sets of the individual “AREAnnnn” files are also available in “zipped” format, collected in 95 self-extracting *.EXE files, with each of these containing all five channels for a 1-minute time period. These zipped, “FAS_A???.EXE” files, can be accessed directly by clicking here.
The following data is from a special 5-minute data set being collected by the new GOES-10 satellite. It covers the period from 0830 to 1335 UTC for the infrared channel data and 1130 to 1335 UTC for the visible channel, and includes the time during which the killer tornado was on the ground. The tornado touched down 2 miles south of Murrayville, GA (about 6 miles north of Gainesville) at 6:20 AM EST and traveled 11.5 miles northeast. 11 people in Hall county, and 2 in White county Georgia were killed.
Our first, quick look suggests the following: All of the precursor action was well before dawn so visible imagery is not helpful during the period when the storm was forming and intensifying. The infrared imagery shows that the storm of interest formed just behind the major squall line, on what may be an outflow boundary. Though the size of the storm is not large, nor is the top very cold (compared to nearby storms), there is a storm-scale circulation evident in the subsidence around the anvil. There is also a possible storm split, with a left moving overshooting top heading off to the north with the synoptic flow, and a right mover moving almost due easterly.
These digital, McIDAS format data files are available by anonymous FTP to: “canopus.cira.colostate.edu”. They are in directory 98079_GA_tor. The 5-channel, five-minute sector files are named with the conventional McIDAS “AREAnnnn” format, with nnnn=1700’s indicating GOES Imager channel 1 (4 km), nnnn=1900’s for channel 1 (1 km), nnn=3000’s for channel 3, nnnn=4000’s for channel 4 and nnnn=6000’s for the fog product. Channel 1 is visible, channel 3 is at 6.7 micrometers (water vapor), and channel 4 is the conventional IR window.
Gator1- Gator50.gif are a series of gif images, centered over Atlanta, showing the enhanced (to highlight the cold cloud-tops) IR imagery from channel 4. These images show the environment well before the storms began, 0515 UTC, through the rest of the period defined above. An 8-frame loop from this dataset can be viewed by clicking on the “thumbnail” image below.
On 3 May 1999, multiple supercell thunderstorms produced many large and damaging tornadoes in central Oklahoma during the late afternoon and evening hours. Some of these storms were killers, including the twisters which moved through and/or near Dover, Shawnee, Perry and Bridge Creek, and the Moore and southern Oklahoma City metropolitan areas. Additional tornadoes also hit areas in south-central Kansas, eastern Oklahoma and northern Texas, with over 50 being observed across the region. The tornado count for this outbreak makes it the largest ever recorded in the state of Oklahoma. The most recent statistics show that 42 people have died as the result of the outbreak and 795 were injured. Many homes and businesses have been destroyed or damaged throughout the affected areas, with a total damage estimate of about $750 million. Five deaths, 100 injuries and heavy property damage were also incurred in the Wichita, Kansas metro area.
The map, courtesy of the NWS-NSSL, shows the approximate location and paths of the most damaging tornadoes. NOTE: This map is based upon a preliminary assessment of the outbreak and it does not include all of the tornadoes known to have occurred within OK. The NWS’ Norman, OK staff is currently in the process of assessing damage and Fujita Scale ratings for the storms.
The available satellite imagery covering this outbreak includes those derived from both GOES-8 and GOES-10, and they span the period from 00 UTC on 3 May to 07 UTC on the 4th. The images are all at the routine operational scan-rate of 15 minutes apart. These digital, McIDAS format data files are available by anonymous FTP to: “canopus.cira.colostate.edu”. They are in directories 99123_OK_tor/Goes_8 and 99123_OK_tor/Goes_10. As in other cases available in the CIRA-RAMM Team Virtual Lab, the imagery is contained in McIDAS area-file format, with AREA1*** containing channel 1 (VIS) images; AREA2***, channel 2; AREA3***, channel 3; AREA4***, channel 4; and AREA5***, channel 5.
For access to severe weather reports, Doppler radar imagery, photographs and other documentation about this storm, visit the Central Atlantic Storm Investigators’ website.
New procedures for requesting super rapid scan operations (SRSO) or one minute imagery from the NOAA geostationary operational environmental satellites in 1999 allowed for requests for data on consecutive days. This produced a very detailed and unprecedented satellite dataset of Hurricane Floyd. Eight days consecutive of two hours or more of SRSO data were collected for Floyd, covering its development from tropical storm to intense hurricane (8 September – 15 September).
Hurricane Floyd acheived its maximum intensity at 12 UTC on 13 September when it was located at 23.6 N, 71.4 W. At this time its maximum sustained winds were 135 knots and its minimum sea level pressure was 921 hPa. Shortly after this time at 1255 UTC the GOES-8 Satellite began SRSO which produces 2 bursts an hour of 1-minute interval imagery, 8 minutes in duration. The satellite stayed in SRSO for the next 3 hours ending at 1542 UTC, producing 5 periods of 1-minute imagery. At this time Floyd had begun to fill but very slowly with central pressures rising to 923 hPa and winds dropping to 125 knots by 18 UTC, where it was located at 24.1 N and 72.9 W.
This is a SRSO (1 minute interval imaging) data set, covering 11 1/2 hours from 10:15 thru 21:55 UTC.
These digital, McIDAS format data files are available by anonymous FTP to: canopus.cira.colostate.edu. They are in directory /99256_floyd. The 5-channel, one-minute sector files are named with the conventional McIDAS “AREAnnnn” format (Approx 1.1 GB).
Also available are a collection of ZIP files for faster download (Approx 530MB).
Channel 1 is visible, channel 2 is at 3.7 micrometers, channel 3 is at 6.7 micrometers (water vapor), channel 4 is at 10.2 micrometers (conventional IR window), and channel 5 is at 11.5 micrometers.
On 24-26 January 2000, the East Coast of the U.S. was covered with snow produced by a low pressure system (central pressure below 980 mb on the 25th) which moved along the coast from South Carolina to New England. Twelve states (NC, SC, VA, WV, MD, PA, VT, NJ, DE, CT, MA, and NH) had reports of 6 or more inches of snow, with a record 20.3 inches falling in Raleigh-Durham, NC. Wind speeds at or above 30 kt were reported at numerous Atlantic buoys. This “Surprise Winter Storm” was not well forecast by the numerical models (see forecast model details), which impeded the efforts of forecasters to stay on top of the situation during its early stages.
For more details about what the imagery can tell us about this major weather event , please visit the Cooperative Institute for Meteorological Satellite Studies’ (CIMSS’) website presentation.
The available GOES-8 dataset consists of some 550 files, covering the period from about 1200 UTC on 24 Jan. through 1245 UTC on 26 Jan. 2000. The file naming convention, yeardayhrmnsei08, can be interpreted as follows: year = 4-digit year; day = 3-digit Julian day; hr,mn, se = 2-digit-each hour, minute and second UTC associated with the image; i, or s as the case may be, = GOES Imager, or Sounder, data; and 08 = GOES-8 data source. The filename extensions denote the associated Imager channel number or, in the case of the Sounder, it is zero (.c00). The Imager channel number 1 (.c01) is the Visible; channel 3 (.c03) , the water vapor; and channel 4 (.c04) , the IR imagery.
These digital, McIDAS-formatted data are available by FTP through “canopus.cira.colostate.edu”. Log in with “anonymous” and then provide your e- mail address as the password; the data are in the /00024_Ecst_Snow/ directory.
This data set covers a supercell storm that occured on July 24, 2000. The storm initiated in southern South Dakota at approximately 20:30 UTC and appeared to split taking a due south course. As it moved across central Nebraska numerous tornadoes, large hail and damaging winds were reported. The thunderstorm maintained supercell characteristics for over 6 hours.
This is a SRSO (1 minute interval imaging) data set covering 3 1/2 hours from 21:30 thru 00:59 UTC.
These digital, McIDAS format data files are available by anonymous FTP to: canopus.cira.colostate.edu. They are in directory /00206_supercell. The 5-channel, one-minute sector files are named with the conventional McIDAS “AREAnnnn” format (Approx 1.1 GB).
Also available are a collection of ZIP files for faster download (Approx 500MB).
Channel 1 is visible, channel 2 is at 3.7 micrometers, channel 3 is at 6.7 micrometers (water vapor), channel 4 is at 10.2 micrometers (conventional IR window), and channel 5 is at 11.5 micrometers.