Polar Hyperspectral Modeling for a narrow line of convection

May 13th, 2022 |
PHSnABI 7-h and 1-h forecasts of CAPE valid at 2300 UTC on 13 May 2022 along with GOES-16 Derived CAPE overlaid with GOES-16 Band 13 Infrared (Band 13, 10.3 µm) imagery, GOES-16 Visible Imagery (Band 2, 0.64 µm) overlain with Radar imagery, and GOES-16 Visible Imagery alone (Click to enlarge)

A narrow ribbon of Slight RIsk was forecast for parts of the midwest on 13 Friday 2022, as shown below, and a few severe weather events occurred (SPC Storm Reports); they were well forecast. How did the Polar Hyperspectral Sounding forecast system perform on this day? The toggle above shows a 7-h forecast of CAPE (initialized at 1600 UTC and valid at 2300 UTC). It’s noteworthy that the forecast also shows a narrow corridor of instability. A similar toggle, but starting with the 0-h initial field of PHSnABI derived CAPE from the model at 2200 UTC, is here.

SPC Day 1 Outlook, 13 May 2022, issued at 2000 UTC (Click to enlarge)

The toggle below shows the 7-h forecast compared to the GOES-16 ABI Derived CAPE. A similar toggle, here, compares the 1-h forecast (initialized at 2200 UTC, valid at 2300 UTC) with the 2300 UTC Derived CAPE observed from GOES. The 7-h forecast below might be too far to the east; however, the developing convection associated with ribbon of instability is removed from the leading edge of the CAPE.

7-h forecast of CAPE from PHSnABI Modeling system and GOES-16 Derived Stability CAPE (overlain with GOES-16 ABI Band 13 Infrared imagery (10.3 µm) at 2300 UTC on 2300 UTC 13 May 2022 (click to enlarge)

Precipitation forecasts from this event (available at this website) are shown below, starting with two forecasts valid at 2300 UTC: a 3-h forecast from 2000 UTC and a 1-h forecast from 2200 UTC. They both show strongest convection over western IL, as observed. The 2000 UTC forecast also shows the break in convection over southern WI, also as observed.

PHSnABI forecasts of 1-h precipitation at 2300 UTC valid from initial times of 2000 and 2200 UTC on 13 May 2022 (Click to enlarge)
Accumulated 1-h precipitation from the PHSnABI model initialized at 2200 UTC on 13 May 2022; forecasts valid at 2300 UTC on 13 May, 0000 and 0100 UTC 14 May 2022 (Click to enlarge)

The PHSnABI modeling system accurately showed the corridor of instability over the Great Lakes, and convection did develop with this instability as observed. (Note: forecasts initialized before 1700 UTC did not produce precipitation; observations from the afternoon overpasses of NOAA-20 and Suomi-NPP perhaps supplied the necessary information leading to a better prediction of precipitation). Radar imagery over WI at 0054 UTC on 14 May 2022 is shown below. The initial (very narrow) line of convection did produce precipitation over Madison, but precipitation moved over Madison from the south after 0100 UTC.

Base Reflectivity at 0054 UTC on 14 May 2022 (click to enlarge)

Hyperspectral forecasting for the Andover KS tornado

April 30th, 2022 |
GOES-16 Clean Window infrared imagery (10.3 µm) from 0001 – 0201 UTC on 30 April 2022 (Click to enlarge)

The animation above shows GOES-16 Band 13 Infrared imagery (10.3 µm) monitoring the development of the tornadic thunderstorm just east of Wichita that spawned the EF-3 Andover, KS, tornado (on the ground from 0110 through 0131 UTC) on 30 April 2022 (SPC Storm Reports; Write-up on storms from NWS Wichita). The storm responsible for the tornado was ahead of the cold front, which front is obvious in the infrared imagery to the west of Cherokee (OK) and Anthony (KS). Polar Hyperspectral Soundings (available at this website) are being demonstrated at the Hazardous Weather Testbed this year (starting in late May); how did those model products do with this isolated storm out in front of the approaching cold front? The animation below shows the Significant Tornado Parameter forecasts at 0100 UTC from forecasts starting from 1800 through 2300 UTC on 29 April. Maximum values of STP are over the southeast Kansas, especially for the forecasts initialized at 2100 and 2300 UTC. Guidance was suggesting where to focus attention on the possibility of strongest convection.

Significant Tornado Parameter forecast valid at 0100 UTC on 30 April 2022 for forecasts initialized between 1800 and 2300 UTC on 29 April 2022 (Click to enlarge)

The PHSnMWnABI model outputs other variables, of course, and the updraft velocity at the Level of Free Convection (LFC) and the hourly Liquid Precipitation are shown below. Note the two separate regions of convection (based on the updraft velocities), and how a strong system is present near Wichita at 0100 UTC from the 2300 UTC initialization. The forecast model is pointing a forecaster towards the type of evolution that might happen: Convection in more than one location — with strong updrafts along the front (notable in the Lifted Index and CAPE fields — not shown) but also convection out ahead of that feature as shown in the updraft velocities in the 2300 UTC model run (shown at bottom)

Updraft velocity at the LFC (left) and accumulated hourly precipitation (right) at 0100 UTC on 30 April 2022 from the forecasts initialized between 2000 UTC and 2300 UTC on 29 April 2022 (Click to enlarge)
Updraft velocity at the LFC (left) and accumulated hourly precipitation (right) at 0100 UTC on 30 April 2022 from the forecast initialized at 2300 UTC on 29 April 2022 (Click to enlarge)

What did other forecasting models show for this event? The animation below shows ensemble mean STP from the HREF initialized at 1200 UTC on 29 April. (imagery courtesy Jim Caruso, SOO at WFO ICT). STP increases in the vicinity of the observed tornado between 0000 and 0100 UTC. HREF Ensemble Mean Most Unstable Convective Available Potential Energy (MUCAPE) has an axis over the Andover region; updraft helicity signatures are also shown.

HREF Ensemble mean values of STP from the forecast initialized at 1200 UTC, valid hourly from 2200 UTC on 29 April through 0200 UTC on 30 April (Click to enlarge)
HREF Ensemble mean values of MUCAPE from the forecast initialized at 1200 UTC, valid hourly from 2200 UTC on 29 April through 0200 UTC on 30 April; Surface-500mb shear is also plotted, as are helicity signatures > 75 (Click to enlarge)

A regression of the 4-h max probability of Tornado (here), from a 2200 UTC initialization and valid through 0200 UTC, shows highest probabilities encompassing most of Butler County (in which Andover is the most populous city). Forecast updraft helicities are greatest just north of Andover.

Polar Hyperspectral Soundings and High Plains Convection

April 29th, 2022 |
PHSnMWnABI CAPE estimates, 1700 UTC 28 April to 0300 UTC 29 April 2022 (Click to enlarge)

On 28 April, SPC’s convective outlook showed a small region of SLGT RSK over western Nebraska, with Marginal probabilities over most of Nebraska and Kansas (link to 2000 UTC outlook). The animation above shows hourly CAPE predictions from a 3-km version of the Rapid Refresh that is run hourly and initialized with Polar Hyperspectral Data (infrared and microwave, from Metop and from Suomi-NPP and NOAA-20) fused with GOES-16 ABI data, thereby using the strengths of ABI (fine spatial and temporal resolution) and Polar Hyperspectral Soundings (excellent spectral resolution). The animation above (a mix of initial fields — 1700 – 1900 UTC ; 1 – 4h forecasts from 1900 UTC; and 6-9h forecasts from 1800 UTC) shows CAPE developing over the High Plains and then rotating north into Kansas by 0300 UTC on 29 April 2022. A later forecast of CAPE, below, runs from 2100 UTC 28 April through 0600 UTC 29 April (showing initial fields at 2100/2200 UTC, then forecasts from 2200 UTC at 2300 UTC through 0400 UTC [i.e., 1-6h forecasts], followed by 8h and 9h forecasts from 2100 UTC on 0500 and 0600 UTC). By 0600 UTC, an axis of instability stretches from western Nebraska southeastward into central Kansas. The two forecasts show similar patterns.

PHSnMWnABI CAPE estimates, 2100 UTC 28 April to 0600 UTC 29 April 2022 (Click to enlarge)

The toggle below compares two forecasts for 0300 UTC, a 9-h forecast from 1800 UTC on 28 April, and a 5-h forecast from 2200 UTC on 28 April. The 5-h forecast is a bit less quick in moving the high CAPE values northward.

5-h and 9-h forecasts valid at 0300 UTC on 29 April 2022 (click to enlarge)

So, what happened? The toggle below compares the 6-h forecast from 2200 UTC, valid at 0400 UTC on 29 April, with the initial field for the 0400 UTC model run. As above, it appears that the forecast model from 2200 UTC was a bit too fast in moving the CAPE northward and eastward into Kansas/Nebraska. But overall there is very good agreement between the two fields.

PHSaABI CAPE at 0400 UTC 29 April 2022: a 6-h forecast from 2200 UTC 28 April 2022, and the initial field for the 0400 UTC model run (Click to enlarge)

The animation below shows PHSnABI CAPE fields hourly from 0400 – 0700 UTC (initial fields from 0400-0600; 1-h forecast at 0700), side by side with observed GOES-16 ABI Band 13 color-enhanced brightness temperatures. The convection that develops is along the edges of the CAPE; that is, it forms along the CAPE gradient in the model. Click here to view SPC Storm Reports from 28-29 April.

PHSnABI CAPE values, 0400-0700 UTC (left) and GOES-16 ABI Band 13 infrared (10.3 µm) imagery, 0401-0701 UTC 29 April 2022 (click to enlarge)

The animation below shows ABI Infrared Imagery overlain on top of the forecast CAPE field. This drives home to point that convection on this day occurred where the gradient of CAPE was outlined/predicted by this forecast model.

PHSnABI CAPE fields overlain with GOES-16 Infrared ABI Band 13 (10.3 µm) imagery, 0800-1200 UTC on 29 April 2022 (Click to enlarge)

PHSnABI data will be demonstrated at the Hazardous Weather Testbed in late May/Early June. Model output is available outside of AWIPS at this website. To view more blog posts on this project, click the ‘Hyperspectral’ tag below.

Polar Hyperspectral modeling during severe weather: Iowa on 12 April 2022

April 13th, 2022 |

Previous posts on the CIMSS Blog (here, here, here) have detailed a modeling system that incorporates thermodynamic information from hyperspectral soundings (CrIS on Suomi/NPP and NOAA-20; IASI on Metop-B/Metop-C) into a version of the Rapid Refresh model. The improved definition of thermodynamic distributions by the hyperspectral data (that is: better spectral resolution) is combined with the better spatial and temporal resolution of the Advanced Baseline Imager (ABI) on GOES-16 to produce 18-h forecasts every hour (model output is available here; a Journal Article describing this data fusion is here). This product is to be demonstrated at the Hazardous Weather Testbed, starting in May of 2022.

Modeled Convective Available Potential Energy (CAPE), 1600 UTC 12 April – 0000 UTC 13 April 2022 (Click to enlarge)

Severe weather occurred in the Plains on 12 April (SPC Storm Reports). What did this modeling system show? The animation above shows Convective Available Potential Energy, hourly from 1600 – 0000 UTC on the 12th. The first two frames are analyses, the final frames are part of the 1700 UTC forecast, which should include some IASI input from the morning Metop overpasses (MetopB orbits on 12 April; MetopC orbits on 12 April). Two areas of enhanced CAPE are present: one starts near Kansas City and lifts northeastward into southeastern Minnesota and western Wisconsin before eroding; a second, larger region of stronger CAPE moves from Oklahoma/Kansas into western Iowa by 0000 UTC. Was convection associated with this instability? That is shown below: side-by-side animations of GOES-16 Day Cloud Phase Distinction and PHSnABI CAPE values.

Strong convection develops initially over northeast Iowa/southern Minnesota just north of the eastern diagnosed CAPE maximum. A second round of convection develops over western Iowa in association with the diagnosed CAPE maximum there.

GOES-16 Day Cloud Phase Distinction (left) and PHSnABI CAPE Values (right), hourly from 1600 UTC 12 April 2022 to 0000 UTC 13 April 2022 (Click to enlarge)

The model data output are archived at this link. It’s pretty easy to view how forecasts are changing with time as an event approached. For example, the animation below compares forecast CAPE values for 0000 UTC on 13 April — 18-h, 12-h, 9-h and 6-h forecasts from 06, 12, 15, and 18 UTC on 12 April 2022. Note the big change between the forecast initialized at 0600 UTC and the one initialized at 1200 UTC. Between those two times, Suomi-NPP and NOAA-20 will have overflown the central United States and helped the initial fields in the model to represent more accurately the thermodynamics. Forecasts after that concentrate CAPE values over western Iowa.

18-h (from 0600 UTC), 12-h (from 1200 UTC), 9-h (from 1500 UTC) and 6-h (from 1800 UTC) forecasts for MUCAPE valid at 0000 UTC on 13 April 2022 (Click to enlarge)

Changes in the Significant Tornado Parameter (STP) for the same time, from the same model runs, are shown below, and tell the same story. Thermodynamics and wind profiles in the model are most strongly favorable for tornadogenesis over western Iowa.

18-h (from 0600 UTC), 12-h (from 1200 UTC), 9-h (from 1500 UTC) and 6-h (from 1800 UTC) forecasts for Significant Tornado Parameter (STP) valid at 0000 UTC on 13 April 2022 (Click to enlarge)

What did the model precipitation look like? CAPE and STP show a potential. But the convection has to develop and be influenced by the CAPE and STP. Model precipitation is shown below. Only the 1800 UTC model run shows the strong convection that developed over western Iowa by 2200 UTC — and that precipitation is a bit too far west.

18-h (from 0600 UTC), 12-h (from 1200 UTC), 9-h (from 1500 UTC) and 6-h (from 1800 UTC) forecasts for hourly accumulated precipitation valid at 0000 UTC on 13 April 2022 (Click to enlarge)

There is also a 3-km version of this model available at the model output link. (This one) The higher-resoluation model run (as might be expected) does a far better job in predicting the onset of convection. Output from the 1800 UTC model run is shown below. Initiation and behavior is very well-matched with observations. The Significant Tornado Parameter from this run (bottom) shows that parameter maximizing over northwest/northcentral Iowa between 2100 UTC on the 12th and 0100 UTC on the 13th. The first tornado per SPC Storm reports occurred shortly after 2300 UTC on the 12th.

Hourly Precipitation (in mm) from the 3-km PHSnMWnABI model run initialized at 1800 UTC on 12 April 2022 (Click to enlarge)
Significant Tornado Parameter from the 3-km PHSnMWnABI model run initialized at 1800 UTC on 12 April 2022, hourly output from 1800 UTC 12 April – 0300 UTC 13 April 2022 (Click to enlarge)

This case demonstrates the power of adding Polar Hyperspectral Sounding information into a model run: An accurate forecast of likely tornadic development resulted. The next generation of Geostationary Satellites (GeoXO) are slated to carry a GXS — a Hyperspectral Sounding in geostationary orbit — that could also supply accurate thermodynamic information to numerical models to enhance forecasts. For more on GeoXO, click here.