April 1st storms fool nobody

Two regions of storms traversed the central part of the U.S. on Wednesday, forced by a cold front in IA/NE/KS, and in a region of moist, unstable air ahead of an MCV in the TX/OK/AR/MO area. Marginal and Slight risk SPC outlook regions corresponded with most of the severe hail and wind reports.

NOAA/CIMSS ProbSevere performed well in a number of cases where satellite growth was observed. The statistical model also tends do perform decently in the spring months, when melting levels are lower, and the MRMS MESH values are more accurate (perhaps due to less melting in the lower troposphere).

Fort Worth, TX had to contend with a slow moving storm, in a very unstable, moderately sheared environment (~3000 J/kg MUCAPE and ~18 kts effective bulk shear). While only weak normalized vertical growth rate was observed by GOES-East at 20:28 UTC, the probability shot up from 27% to 61% at 20:30, when a strong vertical growth rate and a moderate glaciation rate were observed. The MESH also increased to almost 0.75". The strong satellite growth was still evident despite some mid-level cloud cover, and in part thanks to the call for rapid-scan operation of GOES-East. The probability fluctuated a little bit as the MESH oscillated, but was first warned at 20:49 UTC. This storm would go on to produce tennis-ball (2.5") hail in the Fort Worth metro.

Further south, another storm popped up in Brown Co., TX, with moderate normalized satellite growth, and very strong MESH (1-1.5"). The first probability ≥ 70% came at 21:02 UTC, and 1" hail was observed at 21:25 UTC near the Brownwood Regional Airport. No warning was issued for this storm.

A small storm initiated quickly around 7pm CDT in northern Arkansas, in a moderately unstable, moderately sheared environment. Very strong satellite growth was observed at about 23:30 UTC, despite some high cirrus cloud from convection to its west. The probability of severe was 90% at 00:04 UTC, with MESH=0.90". The probability then fluctuated (as low as 39%, briefly) with the MESH, before taking off again. The first severe thunderstorm warning was issued at 00:44 UTC, with golf ball hail reported 20 min later.

One question we ask to forecasters is how can/do you use the probability information when it seems to fluctuate a bit. Most forecasters seem to prefer to see some sustained strength before issuing a warning. Though the type of storm environment may change that inclination, especially for environs with extreme instability.

Lastly, here is a loop of convection along the cold front in the central Plains. Moderate vertical growth was observed for these storms, whereas only weak glaciation was evident, owing to the mid-to upper-level cloud cover during initiation. Lead-times of 6, 10, and 0 minutes to the first NWS severe thunderstorm warnings was achieved by ProbSevere for these first three clusters of storms. In this case, ProbSevere may have been a handy tool to increase confidence for warning, and perhaps increase lead-time by a NEXRAD scan or two.

GOES-East IR brightness temperature with MRMS composite reflectivity, ProbSevere, and NWS warnings overlaid. Reflectivity less than 25 dBZ was masked out, to better see the satellite growth in the IR band.

John Cintineo
UW-CIMSS

Severe storms in the southern Plains

A potent, negatively tilted short-wave trough and associated cold frontal passage brought a round of severe storms to the southern Plains yesterday --  the first large batch of severe storms in the CONUS during March 2015.

The environment was highly sheared and exhibited moderate instability (~50 kts effective bulk shear; ~1600 J/kg MUCAPE). The NOAA/CIMSS ProbSevere model is designed to provide forecasters increased confidence and additional lead-time to the initial severe storm hazards. It does this by linking observed satellite growth and NWP environmental information with developing storms on radar. ProbSevere showed a strong normalized vertical growth rate at 2015 UTC, (see below) observed by GOES-East. The glaciation rate was 'weak', probably due to an overlapping cirrus canopy.

Because of the favorable environment and strong satellite growth rate, the probability of severe was 21% at 2036 UTC, while the MRMS MESH was 0.00 in. As MESH began to increase, the probability also increased (57% at 2046 UTC, 90% at 2050 UTC). The first 1.00" hail report was recorded at 2057 UTC in Bourbon Co., KS (11 min lead-time from ≥ 50%, 7 min lead-time from ≥ 90%). The first severe thunderstorm warning was issued at 2100 UTC.

A little later on and a little further south in far southeast KS, another storm exhibited exceptionally strong satellite growth rates (both normalized growth and glaciation rate) at 2045 UTC. While the MESH was about 0.2", the probability of severe was still 45-55%. A warning was issued at 2136 UTC, when the probability of severe was 52%. The probability eventually topped out at 60%, and 1.00" hail was reported in Jasper Co., MO at 2152 UTC.

The highest MESH in this storm was only 0.30 in. This is an example where observed satellite trends can help compensate for a weak radar signal (perhaps due to the low-precip, or LP nature of a storm), and how blending multiple meteorological data sources can increase confidence and lead-time.

This final animation just demonstrates the evolution of ProbSevere values for the line of storms, and their associated warnings.

John Cintineo
UW-CIMSS

Elevated September Storms

Thunderstorms developed in southern Iowa early Monday morning along a warm frontal boundary. The MLCAPE was 0 J/kg, while the MUCAPE across the baroclinic zone was 800-1000 J/kg, indicating that these storms were not rooted at the surface. The effective bulk shear was also considerable, at 45-50 kts. The storm featured below had strong satellite growth rates at 11:15 and 11:30 UTC, which contributed to an elevated probability of severe (44%) when the MRMS MESH was less than 0.5". The ProbSevere model showed probabilities increase to 84% (12:04 UTC) and 96% (12:14 UTC), as the MESH increased to 0.81" and 1.27". This storm was never warned, while one-inch diameter hail was recorded in Appanoose county, IA at 12:55 UTC, 51 minutes after the first probability of severe greater than 80%. This example demonstrates the utility of the ProbSevere model in perhaps giving forecasters a "heads-up" on storms to watch in a set-up where severe weather was not expected.

John Cintineo

UW-CIMSS

NOAA/CIMSS Prob Severe Examples from September 9 2014

A strengthening frontal boundary extended from Lake Superior to the western High Plains on Tuesday, September 9th, while a surface warm front extended east from developing low pressure in Kansas east to central Illinois.  A vigorous mid-level short wave was moving across western and central Nebraska at the time resulting in the strengthening low pressure and cyclogenesis.  Deep column moisture was located to the south of the warm front with surface dewpoints in the lower 70s and precipitable water values of 1.5 to 2 inches.  Moderate to strong instability was in the frontal zone along with moderate effective bulk shear.  

Several strong thunderstorms developed during the early afternoon in the vicinity of the warm front. A line of thunderstorms moved across southwest Iowa toward Des Moines during the early afternoon. The below image from 1816z shows this line of thunderstorms to the southwest of Des Moines.  The CIMSS Prob Severe product was indicating a 69 percent chance for this line of thunderstorms to become severe in the following 60 minutes.  The thick cirrus anvil from this thunderstorm prevented the Vertical Growth and Glaciation Rate information from being determined.  

An examination of the KDMX all-tilts products and digital VIL indicated more of a threat of strong winds over large hail.  However RAP forecast soundings showed a strong thermal inversion around 1500 feet.  The KDMX velocity imagery indicated approaching 50 to 60 knot winds approaching Des Moines along the radial.  However, the CIMSS Prob Severe product lowered the severe probability to around 30 percent as the line of storms approached Des Moines.  A severe thunderstorm warning was issued for the counties to the west of Des Moines. The thunderstorms moved through the Des Moines area around 1915z uneventfully as the inversion prevented the strong winds aloft from mixing down to the surface. 

Another batch of thunderstorms strengthened over eastern Nebraska during the early afternoon. These thunderstorms approached the Omaha area and produced a wind gust to 55 knots at Fremont Nebraska (KFET) at 1955z which is about 30 miles west of KOMA.  The CIMSS Prob Severe product from 1948z indicated only an 8 percent chance for this thunderstorm to become severe in the next 60 minutes.  Again, the Growth and Glaciation Rate information was not available.  

The above two examples illustrate how the CIMSS Prob Severe product is geared more toward forecasting a large hail threat using available instability parameters, especially when satellite information is not available.

Marc Kavinsky
NWS Milwaukee/Sullivan

John Cintineo
UW-CIMSS

Upper Midwest storms and NOAA/CIMSS Prob Severe Model Demonstration

The GOES-R PG CIMSS/MKX shift focused on the Upper Mississippi Valley on Thursday, September 4.  Ongoing convection existed over far northern Wisconsin early in the morning on Tuesday.  The storms were discrete, elevated supercells that eventually congealed into a linear convective system.  While the NOAA/CIMSS ProbSevere model is designed to increase lead-time to the initial severe warning/hazards, this case demonstrates what others have noted during the spring and summer this year, that the ProbSevere model performs well with discrete mature convection.  Figure 1 below is shows the Green Bay 0.5 degree reflectivity with NOAA/CIMSS ProbSevere model output in the shaded contours around storm cells valid at 1452 UTC 4 September 2014.  These two storms with high ProbSevere values (> 90%) produced golf ball sized hail around this time.

Figure 1.  KGRB 0.5 degree reflectivity and NOAA/CIMSS ProbSevere output valid at 1452 UTC 4 September 2014.

Not shown was the progression of these storms across northern Wisconsin in the morning hours on Thursday.  During this time the ProbSevere values were consistently high (> 85%) with sporadic severe reports across the rural Northwoods.

Additional elevated storms developed over western and southwestern Wisconsin during the early afternoon hours, fueled by warm air advection.  These storms struggled to develop as the atmosphere was strongly capped.  The ProbSevere values associated with these storms remained (correctly) quite low over southwestern Wisconsin.  The storms over west-central Wisconsin did exhibit higher probabilities, as much as ~60% (largely due to much higher values of effective bulk shear and MRMS MESH values approaching 0.75"), but as of this writing have yet to produce severe weather.

The strong cap in place is likely to inhibit surfaced based convection along the cold front in Minnesota through the duration of the shift.

-Justin Sieglaff (UW/CIMSS)
-Ben Herzog (NWS MKX)

FROPA convection on the Plains

Strong early September convection fired on a trailing cold front in the late evening and overnight hours in northern OK, southern KS, and southwest MO. The environment near and along the front was characterized by MUCAPE ~2500-4000 J/kg and effective shear ~20-35 kts. The IR-derived satellite growth rates, as captured by the NOAA/CIMSS ProbSevere model were very strong, during this period of GOES-East rapid-scan operation.

GOES-East IR brightness temperature, 5-minute NLDN lightning plot, and METAR station plots.

 The explosive growth rates, along with strong MRMS MESH and a favorable environment, led to very high probabilities of severe (80-100%) before severe weather was reported from many of these storms.

ProbSevere contours overlaid MRMS reflectivity and NWS warnings.

 The first storm to initiate in south-central KS (shown below) went from 10% (22:30Z) to 57% (22:34Z) probability of severe as the observed satellite growth went from moderate/weak to very strong, while the MESH remained constant at 0.25". Six minutes later, the probability of severe was 79% (MESH still < 0.5"). By 22:52Z, the probability of severe exceeded 90%, as the MESH was nearly 0.75". The probability maxed-out at 100% at 23:14Z, when the MESH was 1.47" and the first 1" report was also recorded at this time. This storm would go on to produce a large tornado near the town of Cedar Vale, in Chautauqua county, as well as baseball-sized hail in Winfield, KS.

The storms in the region produced numerous severe weather reports, including hail up to the size of baseballs, wind gusts up to 80 mph, and large tornadoes. The satellite growth rates from the model may increase forecaster confidence in issuing warnings and perhaps increase lead-times to initial severe hazards.

John Cintineo
UW-CIMSS

Severe Hail in Texas Panhandle

All day we were watching weak convection along a cold front in the Central Plains and ProbSevere as expected with these weak storms was no higher than 5 percent. Later in the afternoon, a few isolated cells in the Texas Panhandle rapidly developed. One storm, the Clarendon storm, became severe within an hour of initiation. We utilized ProbSevere as guidance in determining which storms to focus on for potentially severe hail. ProbSevere made it easy to see the evolution and intensification of the individual cells. Eight minutes before the Clarendon warning, ProbSevere was 43%, two minutes later it was up to 72%, another 4 minutes later it was up to 96%. At warning issuance another 2 minutes later it was 98%. ProbSevere performed well with regards to hail size. During this 8 minutes the MESH increased from .71 to 1.8 inches. After the warning was issued, it continued to increase to 2.4 inches. Local storm reports of 1 to 2.5 inch diameter hail were received.

Nearby cells that were developing around this time had a ProbSevere of 80-98%, however none of these cells had a warning except for the Clarendon storm during the initial time period studied. A half hour later the warning over the Clarendon storm was expanded to include most of the nearby cells that ProbSevere had high probabilities for earlier. There were no storm reports issued as of 15 minutes into the warning for the other cells but regardless there was a warning consistent with the ProbSevere guidance.

Sarah Marquardt and Sean Miller
NWS Milwaukee/Sullivan

Crop damage in South Dakota

The morning of August 14th brought a lone hailstorm to central South Dakota. The NOAA/CIMSS ProbSevere model  did not detect cloud growth from satellite, as the convection had been ongoing for some time. The MUCAPE and effective shear of the environment were also rather weak. However, the MRMS MESH product reflected the development of the storm, increasing from 0.71" --> 0.97" --> 1.19" --> 1.69" over the span of 6 minutes (14:50 UTC to 14:56 UTC). The probability of severe increased from 14% to 80% during that short time. The MESH soon after was in the 2-3" range (likely an overestimate, possibly owing to single-radar coverage), and hail of up to 1.25" diameter was reported to damage the local corn crop at 15:05 UTC. The model provided 9 minutes of lead-time to the initial hail reports (measured from the 80% threshold). This example demonstrates how crucial radar detection/diagnosis of storms is in the ProbSevere model, as well as to forecasters.

ProbSevere contours overlaid MRMS Merged Composite Reflectivity, NWS warnings, and local storm reports.

John Cintineo
UW-CIMSS

Satellite growth rate utility in ProbSevere

Convection fired in central Texas along a weak cold front on the afternoon of Monday, August 11th, in an environment characterized by very weak effective bulk shear (< 5 kts) and strong MUCAPE (2800-3600 J/kg).

In the first featured storm, the normalized satellite vertical growth rate went from 'moderate' to 'strong', and the storm-top glaciation rate when form 'weak' to 'moderate' at 17:22 UTC, as the probability of severe jumped from 18% to 61%, with the MESH remaining constant at 0.67 in. The probability then fluctuated between 40% and 60% before the storm was warned at 17:44 UTC with a probability of severe of 58% and MESH = 0.65 in. The fluctuation may have been a result of very low shear and therefore a more "pulse" nature of convection. The satellite growth rates in this storm signaled enhanced probability of severe weather, whereas the shear and MESH were poor to mediocre. This storm was warned again at 18:39 UTC and attained a maximum probability of over 80%. Thus far, there have been no LSRs for this storm, located about 100 miles southeast of Dallas-Fort Worth, TX.

The second featured storm to the southwest had strong normalized satellite vertical growth and glaciation rates at 17:45 UTC. The probability of severe increased from 28% to 72% to 92% from 17:58 UTC -> 18:00 UTC -> 18:08 UTC, as MESH also increased from 0.46 in. to 1.13 in. over that time. This storm was warned at 18:16 UTC and had a severe wind report at 18:18 UTC (trees blocking intersections).

The third storm featured to the northeast of the first storm also had strong satellite growth rates (it was under the same parent satellite cloud-object as the first storm) before rapidly attaining a probability of severe of 86% at 18:56 UTC (MESH = 0.9 in.). It was severe-warned at 19:03 UTC and had a 1 in. hail report at 19:10 UTC.

These examples show how temporal trends in satellite-derived fields can help signal severe potential in slowly developing and more rapidly developing convection. By integrating satellite parameters (when applicable) with radar and NWP fields, we hope to paint a more accurate picture of storm evolution and initial potential severity.

ProbSevere contours around MRMS merged composite reflectivity.

EDIT: Several of these storms did have marginal wind and hail reports.

John Cintineo

UW-CIMSS

ProbSevere in SE Missouri

A high CAPE, low shear environment fueled convective development in southeastern Missouri during the early to mid afternoon on 6 August 2014. Rapid convective development was captured using the ProbSevere function as can be seen in this image.

Convective development over SE MO at 1936Z. ProbSevere values of first warned storm are shown. 0.5 degree reflectivity was captured from KLSX.

This image shows the visible satellite imagery of the same storms in southeastern Missouri. The storm had plenty of lightning associated with it as it tracked to the southeast.

Visible satellite imagery of SE MO. Note the numerous lightning strikes over this area.

At 1936Z, the severe probability was nearly 100% for the northeastern most cell. The MESH value was 1.84. The high CAPE environment allowed these thunderstorms to develop quickly (storms had moderate to strong satellite growth rates), and ProbSevere did a good job in diagnosing the cell as ProbSevere had severe probability values jump from 26% at 1930Z to 81% at 1932Z. Shortly after this spike, the NWS in Paducah issued a Severe Thunderstorm Warning at 1936Z.

At the time of this post, there were no severe storm reports that came out of this cell. This might be contributed to the environment not being as favorable for severe weather (the area was in a See Text severe category from the Storm Prediction Center). SPC did issue an MD for the area for sporadic microbursts shortly after the initial warning, and there were a few penny sized hail reports for a cell that developed over Poplar Bluff. The ProbSevere MESH value for that cell was ~1.14, with a ProbSevere value of 90%. The ProbSevere value was at least 90% for that cell for about five minutes prior to the first report of penny sized hail.

Schultz NWS-MKX
Cintineo UW-CIMSS