ProbSevere results over tidewater Virginia

May 6th, 2014 |

NOAA/CIMSS ProbSevere superimposed on MRMS radar display over southeastern Virginia. Times as indicated. (Click to enlarge)


The Hazardous Weather Testbed (HWT) exercise (Click here for the HWT blog) is ongoing at the Storm Prediction Center.  One of the new products being tested by forecasters is the NOAA/CIMSS ProbSevere product. ProbSevere in the animation above highlighted a cell that produced hail. The AWIPS-2 readout suggests strong vertical growth, and strong glaciation, at 0215 UTC. (The HWT Blog entry on this storm is here) What did the satellite view?


GOES-13 Visible Imagery (0.63 µm), times as indicated. (Click to enlarge)

Visible imagery, above, from just before sunset, shows nascent convective development east of Lynchburg over southeastern Virginia, and also older convection over the Chesapeake Bay and Delmarva Peninsula. The infrared imagery (10.7 µm), below, shows rapid development of convection over southeastern Virginia after 0000 UTC. The first convective cell, which cell is east of the Outer Banks of North Carolina at 0315 UTC, had cloud-tops that cooled about 12 C in 17 minutes (between 0115 and 0132 UTC); the storm that produced hail, and was warned, had cloud-tops that cooled 20 C in 13 minutes, between 0202 UTC and 0215 UTC. This strong vertical growth contributes to a big increase in the ProbSevere value.


GOES-13 Infrared Imagery (10.7 µm), times as indicated. (Click to enlarge)

When interpreting the radar and satellite imagery, be aware of the effects of parallax on the satellite imagery. GOES-13 imagery displayed here is not corrected for parallax. GOES-13 data are parallax-corrected when used in ProbSevere computations, of course.

NOAA/CIMSS ProbSevere Performance during Tornadic Outbreak

April 27th, 2014 |

Severe weather occurred over the southern Plains/lower Mississippi Valley on Sunday 27 April 2014 (SPC storm reports | GOES-13 IR image animation). How did the NOAA/CIMSS ProbSevere product perform with these storms? Three examples are presented below.

NOAA/CIMSS ProbSevere estimates the likelihood that a growing convective storm will first produce severe weather within the next sixty minutes. The Product uses Rapid Refresh model environmental parameters (Most Unstable CAPE, and Environmental Shear), satellite (GOES-13 only, at present) observations of cloud growth and glaciation, and MRMS radar estimates of Maximum Expected Size of Hail (MESH). A convective tower that grows rapidly (as observed by satellite), for example, will be more likely to spawn severe weather in the next 60 minutes than one that grows more slowly. Similarly, as radar intensities increase, so too do the probabilities. The goal of this product is to increase the lead time for a warning by up to several radar scans.

MRMS Radar and ProbSevere readouts, 21:10, 21:22 and 22:08 UTC on 27 April 2014 [Click to enlarge]

MRMS Radar and ProbSevere readouts, 21:10, 21:22 and 22:08 UTC on 27 April 2014 [Click to enlarge]

The image above shows the radar at 21:10, 21:22 and 22:08 UTC, and also values used in the computation of the ProbSevere product, for a hail-producing storm just south of Dallas. The readout shows the most unstable CAPE and Environment Shear (averaged within the radar object that is outlined), the maximum satellite growth and the maximum glaciation rate that the radar object experienced, and also the MESH. In this case, MUCAPE is 4300-4400 J/kg and Environmental Shear is around 50 kts. The maximum satellite growth for the object being tracked is 1%/minute (% meaning percent of the depth of the troposphere) measured at 20:15 UTC (a time when the radar may or may not have been detecting the developing storm). This moderate growth rate stays attached to this growing convective feature at later times. The ProbSevere exceeded 50% at 21:10 UTC, the first warning was issued at 21:22 UTC and the first report of 1″ hail was at 21:45 UTC. (This storm later produced baseball-sized hail). The storm had moderate growth rates, but very large values of MUCAPE and Shear enhanced the probabilities. Glaciation Rate is reported as N/A, which typically means development under a pre-existing cirrus shield.  That was the case over Dallas, as shown in this animation of visible imagery from GOES-13.

As above, but at 21:20, 21:30 and 22:14 UTC 27 April 2014 [Click to enlarge]

As above, but at 21:20, 21:30 and 22:14 UTC 27 April 2014 [Click to enlarge]

The storm above, over northwestern Mississippi, showed strong satellite growth rates (2.4% per minute) and strong glaciation rates. Probabilities exceeded 50% at 21:14 UTC (just before the 21:20 UTC image shown above) and the first warning was issued at 21:30 UTC, with the first report at 22:31 UTC. This storm produced hail, wind and a tornado.

As above, but at 23:00, 23:08, 23:20 and 23:26 UTC 27 April 2014 [Click to enlarge]

As above, but at 23:00, 23:08, 23:20 and 23:26 UTC 27 April 2014 [Click to enlarge]

A third storm that affected the Kansas City metropolitan area is shown above. ProbSevere exceeded 50% at 23:00 UTC, the first warning was issued at 23:26 UTC and the first hail report occurred 23:31 UTC. Satellite growth for this storm is initially strong (1.5% per minute at 21:25 UTC), but the tracked object linked to that strong growth eventually were lost (Satellite Growth values were lost at 23:20 UTC) and probabilities decrease slightly; they remain relatively high because of the favorable environment and radar observations.

For all three examples above, the probabilities were highest with the convective system that produced the severe weather. The combination of the three components (Rapid Refresh, Radar, and Satellite) is key to the probability. Each individual component has strengths but the combination of predictors is what yields a skillful model with meaningful lead-time.

Severe weather continued on April 28th, including fatal storms in Mississippi.  The image below, however, is for a thunderstorm that formed in Missouri outside the region of SPC’s Slight Risk issued at 1300 UTC, and near the edge of the Slight Risk issued at 1630 UTC.

As above, but at 17:20, 17:22, 17:26 and 17:38 UTC 28 April 2014 [Click to enlarge]

As above, but at 17:20, 17:22, 17:26 and 17:38 UTC 28 April 2014 [Click to enlarge]

The ProbSevere for the developing cell is 24% at 17:20 UTC, with Strong Normalized Vertical Growth Rate (2.3%/min) and weak glaciation (0.01/min) observed at 1655 UTC. At 17:22 UTC, new satellite observations have been incorporated (from the 17:15 UTC scan): the Normalized Vertical Growth Rate is now 3.4%/min; in addition, Glaciation (0.17/Min) is also now strong. As a result, ProbSevere increases to 73%. The Severe Thunderstorm Warning was issued 15 minutes later, at 17:37 UTC, 2 minutes after the first report of severe hail with this storm (in Holts Summit MO, in Callaway County).


The ProbSevere product will be evaluated at the Hazardous Weather Testbed to be held in Norman OK in May.


Unusual October Tornado in Wisconsin

October 7th, 2017 |

GOES-16 ABI Visible (0.64 µm) Imagery, 2042-2227 UTC on 7 October 2017 (Click to animate)

GOES-16 data posted on this page are preliminary, non-operational and are undergoing testing

October is typically not a month when tornadoes are expected in Wisconsin. (And in fact, the month ranks fourth in the numbers of fewest tornadoes nationwide). Nevertheless, a brief tornado occurred on the east side of Madison WI late in the afternoon on 7 October (NWS Milwaukee/Sullivan). The visible animation, above, from GOES-16, shows a thin line of weak convection moving through the central part of the state. A slower animation centered on the reported time of the tornado (2200 UTC), is below. The dynamic state of the thunderstorm tops is apparent, and the 5-minute time-step may not be sufficient to resolve the small time-scale changes in the tornadic storm (the two GOES-16 Mesoscale sectors that provide 1-minute imagery were covering the pending landfall of Hurricane Nate at this time, and were giving severe weather support to Puerto Rico, radar-less in the wake of Hurricane Maria).

GOES-16 ABI Visible (0.64 µm) Imagery, 2152-2222 UTC on 7 October 2017 (Click to enlarge)

The Clean Window Channel, below, with the default AWIPS enhancement, shows little cloud-top structure.

GOES-16 ABI Infrared (10.3 µm) Imagery, 1917-2302 UTC on 7 October 2017, default enhancement (Click to animate)

Coldest cloud-top brightness temperatures with this event were in the -15º to -20º C range, and the default enhancement (with a range from -109º to 55º) showed little gradation in that range. By changing the coldest temperature in the enhancement from -109º C to -43º C, however, a structure in the cold clouds emerged, as shown below. Do not be afraid to change enhancements!

GOES-16 ABI Infrared (10.3 µm) Imagery, 1917-2302 UTC on 7 October 2017, adjusted enhancement (Click to animate)

The NOAA/CIMSS ProbSevere products for this event is shown below (from this site). ProbTor values were negligible; however, ProbWind values for the tornadic cell were around 20-24% leading up to the event. In addition, the cell that produced the tornado had the highest probabilities in the identified cells along the line of convection. The single exception occurred at 2235 UTC, after the severe weather event, when the cell to the north briefly (for only five minutes) showed Probabilities exceeding 40%.

NOAA/CIMSS ProbTor, 2130-2235 UTC on 7 October 2017 (Click to enlarge)

============================== Added =============================

The National Weather Service in Milwaukee/Sullivan WI determined that the tornado was an EF-0 based on a damage assessment. They provided radar imagery at the start of the tornado, during the tornado, at its end, and after dissipation.

Lightning at a Football Game

September 29th, 2017 |

GOES-16 ABI Channel 13 (“Clean Window”, 10.3 µm) Infrared Imagery, 2202 UTC 28 September to 0302 UTC 29 September 2017 (Click to animate)

GOES-16 data posted on this page are preliminary, non-operational and are undergoing testing

The National Football League football game between the destined-to-win Green Bay Packers and the woeful Chicago Bears was interrupted by lightning at the end of the First Quarter on Thursday 29 September 2017. The animation of GOES-16 “Clean Window” 10.3 µm infrared imagery, above, shows a cold front passing easily through the area (in much the same way that Aaron Rodgers passed through the Bears Defense). The slow (slow, but not as slow as the Bears’ Offense) animation, below, shows the coldest cloud top moving over Green Bay at around 0130 UTC. Note that there is a parallax shift in this image: the storm top is displayed north of its actual ground location.

GOES-16 ABI Channel 13 (“Clean Window”, 10.3 µm) Infrared Imagery, 2202 UTC 28 September to 0302 UTC 29 September 2017 (Click to animate)

The UW CIMSS ProbSevere product (from here, Click here for an updated version that includes ProbTor/ProbWind/ProbHail) includes Total Lightning (from the ground network) as one of its predictors, and the stepped animation below, showing 0100, 0115 and 0130 UTC, highlights the lightning-producing cell that delayed the game. The Probability of Severe weather (Severe being defined as a Tornado, Hail exceeding 1″ in Diameter, and/or winds exceeding 50 knots) was very small (though not as small as the Bears’ chances), but note that lightning flashes, 2-3 flashes per minute, are consistently observed.

NOAA/CIMSS ProbSevere, 0100, 0115 and 0130 UTC on 29 September 2017 (Click to enlarge)

A new-to-operations instrument that is on board GOES-16 is the Geostationary Lightning Mapper (GLM). Preliminary observations from this instrument, below, show that it too detected the lightning as it approached the stadium. The GLM pixel size is 8 kilometers at the sub-satellite point, a pixel size that is significantly larger than the GOES-16 ABI Pixel size (10.3 µm, with a pixel size of 2 km at the sub-satellite point, is shown below). Whereas ground-based lightning detection systems detect only cloud-to-ground lightning, the optical detectors on GLM detect both cloud-to-ground and in-cloud lightning. Typically, a cloud will produce in-cloud lightning before cloud-to-ground, so the GLM can alert a forecaster to potentially dangerous lightning with more lead-time than is possible with ground-based systems. The animation below starts with the first detection at 0030 UTC on 29 September, about 45 minutes before Lambeau Field was evacuated, and 10 minutes before ground-based sensors detected cloud-to-ground strokes. (the evacuation occurred at about 0115 UTC) Group Density is plotted on top of the GOES-16 10.3 µm ABI (the ABI has a grey-scale enhancement). Note the relatively large pixel size of the GLM, and the obvious parallax shift between the two fields. The strengths of the GLM for lightning safety at large outdoor events is obvious in this case.

GOES-16 ABI “Clean Window” (10.3 µm) Imagery and GLM Group Density, every 5 minutes from 0030-0130 UTC on 29 September 2017 (Click to animate)

Go Pack!!