Polar Hypsectral Sounding and ABI (PHSnABI) Model results during Week 1 of the HWT

May 27th, 2022 |

One of the products being demonstrated at Hazardous Weather Testbed the week of 23-27 May 2022 was a modeling suite that includes Polar Hyperspectral Soundings (PHS, using IASI/AMSU data from the MetopB and MetopC Satellites as well as CrIS/ATMS data Suomi/NPP and NOAA-20) that are associated with Advanced Baseline Imager (ABI) information on GOES-16 through a process known as Data Fusion (or PHSnMWnABI). Previous blog posts on PHSnMWnABI modeling can be viewed here. The summary slides below (courtesy Bill Smith, Sr) show summary results from the four days, including Storm reports from 23 May, 24 May, 25 May and 26 May. Some of the slides are followed by AWIPS screen captures from the various days. Forecasters at HWT were typically active from 1800-2300 UTC on each day.

Monday 23 May 2022

The first day was a day to get one’s feet wet. The WFOs chosen on this day were WFO CAE and WFO LBB. (Here’s the 1630 UTC Convective Outlook for the 23rd; Mesoscale discussions 879, 880, 881, 882, 883, 885, 886 discussed the environment in/around WFO CAE between 1700 and 2100 UTC; Mesoscale discussion 884 focused on the High Plains (CIMSS Blog post on the High Plains convection).

SPC Storm Reports from 23 May 2022, along with Significant Tornado Parameter predictions from PHSnABI (left) and HRRR (right) valid at 2200 UTC (Click to enlarge); HWT CWAs on this day were WFO LBB and WFO CAE.
ProbSevere (Version 3) polygons superimposed on top of a 2-hour forecast of Most Unstable CAPE, all valid at 2200 UTC on 23 May 2022 (Click to enlarge)

A consistent feature of PHSnABI model output of instability over the course of the week was the accurate depiction of instability corridors, as shown above. Invariably, convection formed along the edge of those corridors, on the instability gradient.

Tuesday 24 May 2022

Without doubt this was the most active day of for HWT during this week (CIMSS Blog Post). Forecast offices were WFO MAF, WFO SJT and WFO FWD, a string of offices from the High Plains of Texas to north-central Texas. (Here’s the 1630 UTC Convective Outlook from 24 May.) The summary slide for this day is below. The 3 WFOs chosen this day are aligned with the corridor of stronger SigTor over the high plains of central Texas.

SPC Storm Reports from 24 May 2022, along with Most Unstable CAPE predictions at 1500 UTC (left) and SigTor predictions from PHSnABI, forecast times as indicated (right) (Click to enlarge); HWT CWAs on this day were WFO MAF, SJT and FWD.

For much of this day, I was watching what the forecasters assigned to the San Angelo WFO looked at. The image below shows the Day Cloud Phase Distinction as the strongest storm was starting to form near Tom Green County in west Texas (the country with the odd panhandle!). ProbSevere LightningCast nicely outlines the regions where initiation is ongoing; note the highest values are in a region with lightning observations. That strongest cell is at the intersection of two boundaries, one from southwest to northeast through the developing cell, and one extending east-northeastward from the cell (there’s a prominent wind-shift from southwest to northeast across the line).

Day Cloud Phase Distinction at 2000 UTC on 24 May 2022, along with GLM Observations of Flash Extent Density, and contours of ProbSevere LightningCast (Click to enlarge)

The two-hour forecast from PHSnABI, below, shows largest STP very close to where the convection is shown to be forming above. Note also how STP extends to the east-northeast along the boundary. STP more than once gave a 2-h forecast showing largest values very close to where convection developed, as in this case.

PHSnABI 2-h forecast of Significant Tornado Parameter valid at 2000 UTC on 24 May 2022 (click to enlarge)
3-h forecast of SigTor from PHSnABI, valid at 2100 UTC on 24 May 2022, along with ProbSevere object polygons valid at the same time (click to enlarge)

The largest hail (grapefruit-sized) was reported at 2118 UTC in Tom Green County, shortly after the image above. As on the 23 May, ProbSevere contours are aligned along the PHSnABI parameter, in this case SigTor. The image below shows SigTor predicted values enhanced along the boundary to the east-northeast of the strong storm that by 2200 UTC was east of Tom Green County.

GOES-16 Day Cloud Phase Distinction from 2200 UTC on 24 May 2022, and a 2-h forecast SigTor valid at 2200 UTC (Click to enlarge)

Gridded NUCAPS estimates of 850-mb temperature for this event over the high plains of Texas are shown below. 850-mb temperature north of Tom Green County are around 17oC; to the south they are closer to 23oC: this was a region of warm air advection.

850-mb Temperature fields, gridded from NOAA-20 NUCAPS observations, 1930 UTC on 24 May 2022 (click to enlarge)

Wednesday, 25 May 2022

The summary slide for this date, a much quieter day than 24 May 2022, is shown before (1630 UTC Convective Outlook). WFOs chosen on this day were WFO IWX, WFO MEG and WFO BHM. The tornado on this day (Storm Reports; summary from WFO MKX) was an EF-0 over southern Wisconsin, and SigTor at 2000 UTC on 25 May 2022, and a 4-h forecast valid at 0000 UTC on 26 May 2022 are shown below.

SPC Storm Reports from 25 May 2022, and SIgTor initial field (2000 UTC on 25 May) and the 4-h forecast, valid at 0000 UTC on 26 May (Click to enlarge)

I spent a lot of time on this day watching ProbSevere LightningCast fields over WFO BMX for a simulated Decision Support event related to lightning that is detailed by the NWS forecaster participant here. LightningCast probabilities over Lake Lurlene stayed at/below 50% while western convection moved around the event. Finally, convection moved in from the southwest. An accurate estimate of lightning offset occurred. I was also looking at a line of agitated cumulus moving toward Memphis from western Arkansas that ultimately did nothing. There were subtle features in both PHSnABI CAPE fields and SigTor fields that aligned with this line.

Thursday 26 May 2022

The summary slide for this date is shown below. On Thursday (1630 UTC Convective Outlook), three CWAs were chosen: WFO LMK, WFO GSP and WFO ILX.

Storm Reports from 26 May 2022 (left) and SigTor forecasts at 2200 and 2300 UTC on 26 May 2022 (click to enlarge)

I spent the beginning of the exercise with the Lincoln, IL forecasters; that CWA had mostly sunny skies. The toggle below shows the Day Cloud Phase Distinction at 1900 UTC (shortly after we all started looking at the weather) with the PHSnABI predictions of SigTor and CAPE. The PHSnABI forecasts definitely capture the back edge of the convective field over central Iowa. The strongest cell, isolated within the warm air over the eastern part of the ILX CWA, did eventually become warned as it moved along the border Indiana (Here’s a ProbSevere image from 2000 UTC — from this website). The readout of values within the radar-based object (from this website) are here. SPC storm reports show severe wind at 1947 UTC.

GOES-16 Day Cloud Phase Distinction, 1900 UTC on 26 May 2022, PHSnABI 1-h forecast of SIgTor (overlain with ProbSevere shapefile contours) and PHSnABI 1-h forecast of MUCAPE, both forecasts valid at 1900 UTC on 26 May 2022

Animations of both MUCAPE and SigTor from PHSnABI show features lifting northeastward, as observed.

In the Louisville CWA, widespread clouds were present. The PHSnABI estimate for CAPE, below, show a skinny region of CAPE entering the western part of the CWA.

PHSnABI estimates of CAPE from a forecast initialized at 1800 UTC, 00h – 03h forecasts (Click to enlarge)

I was curious about the region of no CAPE that develops in the northwestern part of WFO LMK at 2000 UTC (which apparently moves out of the LMK CWA by 2100 UTC). Is that convection that the PHSnABI has captured? The toggle below compares the GOES-16 Band 13 imagery, regional radar imagery (overlain with ProbSevere contours) and the predicted MUCAPE at 2000 UTC. The region of very small CAPE is very close to the observed radar convection. (It’s not quite so close in the 2100 UTC imagery.) Note that radar convection has a ProbSevere contour (here’s the image from the ProbSeverev3 website, with the readout for the radar object here), but the presentation from the GOES-16 imagery is not eye-catching. ProbSevere can help focus a forecaster’s attention.

Predicted MUCAPE at 2000 UTC (a 2-h forecast initialized at 1800 UTC), GOES-16 Band 13 Imagery, and regional radar valid at 2000 UTC (Click to enlarge)

The Hazardous Weather Testbed continues on 6 June. Kudos to Kevin Thiel, the SPC Satellite Liaison for GOES-R, and to Kristin Calhoun, SPC, for coordinating this event! For more blog posts from HWT, check out this blog!

Supercells in the Southeast

May 6th, 2020 |

A cold front with ample moisture and instability ahead of it spawned numerous strong storms in the Southeast U.S. yesterday; particularly one long-lived supercell in South Carolina. A convolutional neural network model (CNN) was deployed in realtime on the 1-min GOES-16 mesoscale sector imagery. The model produces an “Intense Convection Probability” (ICP). The inputs for the model are the GOES-16 ABI 0.64 µm reflectance, 10.3 µm brightness temperature, and GLM flash extent density. It was trained to identify “intense” convection as humans do, associating features with intense convection such as strong overshooting tops, thermal couplets (“cold-U/V”), above anvil cirrus plumes (AACP), and strong cores of total lightning.

The animation below shows the ICP contours overlaid ABI 0.64 µm + 10.3 µm sandwich imagery, annotated with preliminary severe storm reports.

The long-lived supercell in South Carolina exhibited AACP and cold-U features, and produced numerous severe wind and hail reports (up to the size of tennis balls). While the NOAA/CIMSS ProbSevere models handled this storm well, the ICP ramped up on a couple of severe storms in northern Georgia before ProbSevere did. ICP for these cells exceeded 90% 15-18 min before ProbWind reached 50%. The ICP may be able to provide additional lead time and confidence to ProbSevere guidance for certain storms, utilizing spectral and electrical information from geostationary satellites. Incorporating ICP into ProbSevere is an active area of current research.

ProbSevere storm contours and MRMS MergedReflectivity for storms in GA and SC. The main or “inner” ProbSevere contour is shaded by the probability of any severe weather, while the outer contour is shaded by the probability of tornado, which appeared when that value was at least 3%, in this example.

An accumulation of ProbSevere storm centroids (white to pink squares, 50% --> 100%), NWS severe weather warnings, and SPC severe local storm reports from 12Z on May 5th to 12Z on May 6th [click to enlarge]

An accumulation of ProbSevere storm centroids (white to pink squares, 50% –> 100%), NWS severe weather warnings, and SPC severe local storm reports from 12Z on May 5th to 12Z on May 6th [click to enlarge]

Thunderstorm with abundant hail over southwest Oregon

April 5th, 2020 |

GOES-17 Band 02 (Red Visible, 0.64 µm) from 1500 UTC 5 April to 0300 UTC on 6 April 2020 (Click to play animated gif)

A thunderstorm over Jackson County, Oregon late in the day on 5 April 2020 produced abundant hail that accumulated to depths in excess of an inch over a wide swath of the county.  Visible animation from 1500 UTC on 5 April through 0300 UTC on 6 April show the region experiencing plentiful sun during the day, helping to destabilize the lower troposphere.  Much of the hail fell in the hour between 0030 and 0130 UTC on 6 April.  That time is captured in the visible loop above, and in the rocking animation from 00 to 02 UTC on 6 April 2020 (and back) below.  Similar animations (from 1500 UTC 5 April to 0300 UTC 6 April) are also available for GOES-17 Band 13 (10.3 µm) and the Day Cloud Phase RGB.

GOES-17 Band 02 (Red Visible, 0.64 µm) rocking animation from 2300 UTC/5 April to 0200 UTC/6 April 2020 (Click to enlarge)

The storms formed in a trough at 500 mb that was associated with abnormally cold temperature (image below courtesy Mike Stavish, WFO MFR). This feature is also discussed in this blog post.

500-mb Temperatures at 0000 UTC on 6 April 2020 (Click to enlarge)

The 0000 UTC Upper-Air sounding at Medford, below, (also courtesy Mike Stavish) shows modest Convective Available Potential Energy (CAPE) amounts and a low freezing level (less than a mile above ground) suggest the possibility of hail.

Medford Oregon Sounding from 0000 UTC on 6 April 2020 (Click to enlarge)

Indeed, the storm in question was a prolific hail producers, as shown in photographs below (photos courtesy Dan Weygand, WFO Medford)


NOAA/CIMSS ProbSevere display at 10-minute timesteps from 0000 to 0200 UTC on 6 April 2020 (Click to enlarge)

NOAA/CIMSS ProbSevere readouts in a timeseries, below (courtesy John Cintineo, UW-Madison CIMSS), shows very low values of parameters typically associated with severe weather.  ProbHail peaks at only 2%. The animation above shows very small values (Click here for a ProbSevere image at 0130 UTC with a Probe that shows values).   Such small values during this event amplify the message that local knowledge of conditions that are favorable for severe weather (and hail events like this that coat roads can lead to traffic crashes and fatalities) are important.

NOAA/CIMSS ProbSevere readout for the Radar Object 16917 in County Oregon

Rocking animations that cover the time of hail fall are shown below.  Because the storm formed/decayed as the sun was setting, use of visible imagery (at top) or Day Cloud Phase Distinction (that uses the visible 0.64 µm in the ‘green’ and the snow/ice near-infrared 1.61 µm in the ‘blue’) was a challenge.  Clean window imagery of course maintained a consistent signal through sunset.

Note the Day Cloud Phase Distinction does capture the glaciation of the cloud.  Sometimes this can be used to judge when lightning might become a hazard.  Lightning was a rare event on this day.

Rocking Animation of GOES-17 Clean Window IR (10.3 µm) from 2300 UTC/5 April to 0200 UTC/6 April (and back) (Click to enlarge)

Rocking Animation of GOES-17 Band 13 Clean Window Infrared imagery (10.3 µm) from 0000 UTC to 0200 UTC (and back) on 06 April 2020 (Click to enlarge)

Rocking Animation of GOES-17 Day Cloud Phase Distinction Red-Green-Blue composite imagery from 2300 UTC/5 April to 0200 UTC/6 April (and back) (Click to enlarge)

NOAA-20 overflew the West Coast, and NUCAPS soundings were produced, at 2047 UTC on 5 April.  (Click here to see NOAA-20 orbits on that day, from this site).  Gridded NUCAPS fields (from this site; update: from this site) of 500-mb Temperatures, Ozone Anomalies, and 850-500-mb Lapse Rates, below (imagery courtesy Emily Berndt and Frank LaFontaine, NASA Sport), show the cold air in the cutoff and the unstable environment that supported the convection.

Gridded NUCAPS estimates of 500-mb Temperatures, Total Ozone Anomaly and 850-500 mb lapse rate, data from NOAA-20 afternoon pass on 5 April 2020, ~2047 UTC (Click to enlarge)

(Thanks to Mike Stavish, Science and Operations Officer — SOO — in Medford for alerting us to this case)

Why 1-minute data matters: Beavertails

June 4th, 2015 |

GOES-14 Visible (0.6263 µm) Imagery, 04 June 2015.  1-minute imagery on the left, 5-minute imagery on the right (click to play animation)

GOES-14 Visible (0.6263 µm) Imagery, 04 June 2015. 1-minute imagery on the left, 5-minute imagery on the right (click to play animation)

Beavertails are ephemeral cloud features that form in the inflow of supercell thunderstorms. They are horizontally long and roughly parallel to the inflow warm front. Its appearance (and presence) is affected by and influences inflow into the storm, and by inference, it affects radar returns. That is — a change in the Beavertail cloud can precede a change in radar. Accurate detection of this cloud type, then, aids the understanding of evolving storm morphology. The animation above shows a severe convective system over southeastern Wyoming, viewed at 1-minute intervals (Left) and at 5-minute intervals. Beavertails that form persist for about 30 minutes, so 5-minute imagery will resolve them; however, the resolution of the 1-minute data is far better to monitor the small changes in size and shape that are related to storm inflow.

Do beavertail changes affect the radar? The animation below shows the ProbSevere product readout from 2000-2220 UTC (Courtesy John Cintineo, CIMSS) (Click here for a slow animation). (Click here for an animation (from 1918-2058 UTC) that includes warning polygons). The increases and decreases in the MRMS MESH appear unrelated to the formation/decay of the various beavertails.

NOAA/CIMSS ProbSevere Product, 2000-2020 UTC on 4 June 2015 (click to animate)

NOAA/CIMSS ProbSevere Product, 2000-2020 UTC on 4 June 2015 (click to play animation)

This storm was captured by different chasers. This YouTube video from Scott Longmore shows the evolution of the convective system from the ground. Hat/tip to Jennifer Laflin, NWS EAX and Chad Gravelle, OPG, for alerting us to this case.