Simulated GOES-R band differences

As part of the EFP CI desk's morning forecasts, they asked me to demonstrate the NSSL-WRF simulated 10-12 micron band difference provided to us by CIRA. Neither of these channels are currently available together on our operational GOES satellites and will be available on the GOES-R satellite once it launches. One of the advantages of simulating satellite data from a model is that we have the opportunity to produce channels that we don't have currently, and we take full advantage of this by producing all 9 of the non-solar GOES-R IR bands. The 10 micron channel is a very clean window, and thus is very sensitive to surface temperature. The 12 micron channel however is sensitive to low-level water vapor. As moisture moves into a clear pixel area, the 12 micron brightness temperature will decrease, whereas the 10 micron temperature should stay the same. When this occurs, the 10-12 micron channel difference will become strongly positive and indicates areas of moisture convergence or pooling, which can lead to destabilization and subsequent convective initiation. Below is a collection of today's notable images, with signals of moisture pooling and destabilization shown in yellow and orange colors...

NSSL-WRF simulated 10-12 micron band difference for 1600 UTC (top), 1900 UTC (middle) and 2000 UTC (bottom) on 24 May 2011.
At 1600 UTC (top image above), we can see that the channel difference is showing an area of low clouds (blue/green colors) beginning to dissipate over central Oklahoma. At 1900 UTC (middle image above) these low clouds are completely dissipated and we can start to see some development of pooling moisture along the dryline in W. OK and the triple point on the OK/KS border. By 2000 UTC (bottom image above), storms begin to initiate near the triple point and values in the channel difference become strongly positive just south along the dryline.

NSSL-WRF simulated 10-12 micron band difference for 2200 UTC on 24 May 2011.
At 2200 UTC (image above) additional convection develops on the southern part of the dryline in central OK and into TX. It is interesting to note the presence of linear bands of enhanced moisture pooling where the convection develops ahead of the dryline. It is theorized by the EFP CI desk participants that this may be signals of horizontal convective rolls within the model leading to areas of enhanced convective potential. This demonstrates a very interesting tool to help aid forecasters in the prediction of convective initiation and also a unique combination of satellite and model information.

Examining NSSL-WRF simulated imagery/lightning

Observed (top) and simulated NSSL-WRF (bottom) WV imagery for 1300 UTC on 24 May 2011.
This morning EFP forecasters and scientists examined the simulated satellite imagery and lightning threat products from the NSSL-WRF. One of the most useful aspects of creating simulated satellite imagery from a model is the ability to compare the output directly to observed satellite imagery to determine model performance, as well as being able to have a one-stop 3-D representation of the model produced atmosphere. At 1300 UTC we matched the simulated GOES-R band 9 (6.95 micron) to the observed GOES-13 WV channel to determine model performance for the day. As you see from the images above, the model atmosphere and the observed atmosphere are very similar. The model correctly identifies an MCS over N. KS and on the KY/TN border. It should be noted that the model's simulated cloud tops generally appear to be smaller and less extensive with the cirrus shields. The position of the mid-level jet streak is also very similar, as seen in the deeper red colors extending from AZ/NM/TX panhandle.

NSSL-WRF simulated lightning threat for 2000 UTC (top) and 0000 UTC (bottom) for 24 May 2011
The NSSL-WRF initiates storms between 1900-2000 UTC along the OK/KS border. The simulated lightning threat imagery (seen above) provided by Bill McCaul (USRA) provides us with an estimation of total lightning flashes. At 2000 UTC (top most image), we see the first isolated storm along the border, with total flashes per square kilometer per 5 minutes reaching a value of 5. At 0000 UTC (bottom most image), we can see the extent of the storms across most of OK along the dryline. At this time we can see flash rate values reaching 12 and can start to get an idea of storm tracks.

NSSL-WRF simulated GOES-R low- (top), mid- (center), and high-level (bottom) WV imagery for 0000 UTC on 24 May 2011.
As we continue through time past initiation, we can take a look at the three GOES-R WV channels produced from the NSSL-WRF imagery. Each channel peaks at a different level in the atmosphere, essentially providing us with 3 layers of water vapor measurements. Currently unavailable on our operational GOES satellite, this will be available with GOES-R and we can simulated it using numerical models. What is particularly interesting in this imagery at this forecast time is the presence of a very strong dry signature (red colors) extending from the TX panhandle into central OK in the high-level water vapor (bottom image above). This is an indication of very strong mid-level jet streak. This signature is present in the lower level WV images as well.

High risk today....

Just wanted to quickly post about the weather situation today... attached below are the 1300 UTC day 1 probabilities from the SPC... One forecast sounding analogue for later today is 3 May 1999. May get a little hectic in here today, but we will do our best to keep up with things on the GOES-R front...

1300 UTC Day 1 Outlook
1300 UTC Day 1 Tornado Probabilities
1300 UTC Day 1 Wind Probabilities
1300 UTC Day 1 Hail Probabilities
For more information, please visit http://www.spc.noaa.gov/

SATCAST captures initial CI along dryline

At 1855 today, SATCAST flagged development occurring in Beckham county in Southwest OK (top left panel).



Forecasters have been waiting for the convection to develop along the dryline in West Oklahoma during the afternoon hours. Despite an environment characterized by MLCAPES greater than 3000 J/kg, SATCAST was able to have a lead time of approximately one hour before 35 dBZ reached the -10 C isotherm (bottom left) and lightning as detected by the PGLM (bottom right).


The early storms of the afternoon struggled initially develop, but have started developing rapidly in recent time frames. SATCAST has caught many of the storms which have developed along the dryline this afternoon as GOES is currently operating in rapid scan mode.

Ultimate CI

Visiting scientists and NWS forecasters visiting the EWP this week have begun looking at real-time data with the promise of some exciting weather over OK this evening. We are currently examining the SATCAST and UWCI products in combination with the multi-radar multi-sensor (MRMS) and PGLM products to anticipation convective initiation. The participants have worked together to develop what has been so lovingly called the "ultimate CI" 4-panel display within AWIPS. The 4-panel (shown above) includes the following products that are linked to the nowcast and detection of CI... Starting from top left and moving anti-cyclonic (clockwise) we have the visible with SATCAST, visible with UWCI, visible with MRMS reflectivity at -10 C, and finally visible with PGLM and NLDN lightning detections. The 4-panel has been saved as a procedure that forecasters can now load very quickly within their AWIPS D2D workstations throughout the rest of the week. Future visitors will have this available as well. This helps demonstrate the ability to combine these unique datasets from multiple sensors into one effective decision support tool.

EWP end of week debrief... 20 May

Today we spent a couple hours soliciting the forecasters for feedback on all of the products that they worked with this week within the EWP. Given the opportunity, I asked the forecasters some follow-up questions based on some observations this week and from yesterday's event, as well as some of their survey responses. Below are the comments from the discussion...

Convective Initiation

- The CIMSS product had pretty good lead time before we actually started seeing lightning of about 45 minutes to 1 hour, only about 15 minute lead time over 35 dBZ echo.

- Later in the events everything became cloud masked.

- Mostly masked with the CIMSS stuff, but UAH was not and we were expecting CI behind the initial line, but nothing went and there were no CI nowcasts made, so that was very good that we weren't getting false alarms.

- There were times when they would have 30 minute lead-times on radar echos, and other times there were no lead-times during the same event.

- I would like to spend some time looking at those products in more my type of environments, like weak shear.

- A probabilistic approach might be more useful than a simple yes/no output. I like the idea of having pre-CI through CI ongoing information.

- "I tended to look at the UAH one more because it was giving me more detections."

- "I found having the masking overlaid was very important... there were times where the CIMSS wasn't showing something but the UAH was and it helped me get an idea of why."

Overshooting-top / Thermal Couplet

- Later on in the evening we saw some detections, but earlier on we could see some enhanced-v and OT signatures in the imagery, but none were detected.

- If you had someone in operations during warning times that was just doing mesoanalysis and telling forecasters that a detection occurred would be very useful... I get too involved with interrogating radar data during warnings.

- Rapid-scan, high resolution satellite data would definitely make this more useful.

- Would like to see an overshooting top collapse product... maybe an alert.

- Only had a chance to look at the icon detections.

Nearcast

- Saw a moisture tongue coming through, but nothing happened.

- Focused primarily on theta-e product, trying to figure out best way to use that... the challenge the day before was that nothing happened, so it's hard to find the values that are more significant... didn't notice any strong signals yesterday, so I didn't really use it.

- Were not able to get the multiple levels in AWIPS... would like to see those.

- "My initial thoughts were that this was no different than looking at the RUC theta-e product... but I do understand that it was nice to have it based on the observations."

- Would definitely like to see this in my home WFO.

Pseudo-GLM

- What could be really useful for forecasters is some training on total lightning activity and how it relates to what's going on within storms.

- Specific numbers for what an intense flash rate is would be helpful.

- We're fairly comfortable with looking at CG activity and what that means, but an IC/CG ratio product would be useful.

- You can get a sense for the trends, especially when significant, but there's a lot of complexity when looking at the entire storm with some areas decreasing and increasing rapidly within the same cell. Maybe if you could interrogate the storm and get a graph of the total lightning activity.

- We have a lot of users like golf courses and parks that we may not be paying attention to, especially for lightning safety in stratoform rain regions where people may think that it's just light rain. When I'm in warning operations, I get engaged in the base radar data and pay less attention to everything else. This may help a lot in those situations.

Overall

- If we had it set up that one person was in charge of the warnings and looking at the radar data, then the other person could focused on a couple of the experimental products if you're working in pairs at the same desk. If the other desk could do the same thing with a couple of the other experimental products, we might be able to get more useful information on everything.

- "Maybe if we had a couple hours each day to just focus on one product, so all of the fields in the product could be examined. I know I set up a real quick procedure on the first day and only got to look at a couple of the fields and felt like I left the rest behind."

- "I'm wondering if a 2-week period would be better for each forecaster... the first week could be used to get comfortable with all of the products and then the second week we could really get into the products in warning operations." (Difficult to get a forecaster away from their office for 2-weeks... especially in May.)

- If forecasters had a week or two to go through articulate presentations prior to arriving that would help us hit the ground running on the first day and avoid powerpoint death. Make it a prerequisite then have a short discussion with the PIs on what they want to focus on during the first day and go into short DRTs for each product.

Active CI along dryline continues...

CI continued along the dryline in eastern Oklahoma into southern Kansas continued for several hours on the afternoon of 19 May 2011. The SATCAST CI Nowcasting algorithm provided ~15-45 min lead times for the occurrence of 35 dBZ echoes with several storms. With the lack of cirrus clouds, individual cumulus clouds could be monitored as they developed into (at times) long-lived supercells which produced tornadoes.

Shown is the SATCAST CI nowcast at 2002 UTC (top image, with 2000 UTC surface observations), with the dryline and four CI locations highlighted. Also, a 4-panel radar image set (from 2018-2100 UTC) with the actual CI locations shown, with approximate lead times on CI provided, as subjectively determined. HWT forecasters found these nowcasts valuable as they were able to determine quickly that CI along the dryline was continuing and would be for the coming ~1 hour past the nowcast time of 2002 UTC. The environment on this day was accompanied by ~2500-2700 Jkg^-1 CAPE values, especially in areas further south along the dryline into far northern Texas. The high CAPE values imply a lower lead time on CI in general, as cumulus clouds grow very rapidly between 15-min time resolution images. Having 5-min GOES data will substantially improve our ability to provide more timely CI nowcasts.

Busy CI day in Northeastern U.S.

On 18 May 2011, CI was occurred in many locations in Virginia, West Virginia, far northern North Carolina, and southern Pennsylvania. Shown is the SATCAST CI nowcast at 1730 UTC, with radar images between 1730 and 1817 UTC. Annotated on the SATCAST image are the "CI locations" labeled (a)-(k). On the radar time series, the same regions are highlighted. In the lower right image of the radar sequence, on the 1817 UTC image, are the locations (circled or with an arrow pointing to where the CI event occurred), and the approximate lead times (in minutes) that SATCAST provided. These are subjectively determined. Events where no CI event was observed (a follow-on echo of 35 dBZ intensity) are labeled as "F" for "false alarm."

This event/time period shows how SATCAST performs in a situation in which CI was widespread. It remains challenging for a forecaster to identify exactly where a CI event forms related to a given CI nowcast despite the high number of SATCAST CI nowcasts. In this case, the flow was southerly and hence all CI events occurred to the north (or north-northwest) of a nowcast location. These storms were generally low-topped (25,000-30,000 ft), directly associated with a closed upper low circulation.

Elevated "convective initiation" in a dry environment

One challenge is nowcasting convective initiation when it occurs within environments which are very dry, and/or when the convection is substantially elevated from surface-based heating, leading to convective clouds which fail to produce echoes of 35 dBZ intensity at ground level. In these instances, a CI nowcasting algorithm that identifies this convective development technically fails as the traditional radar-based CI definition of a 35 dBZ echo (whether at ground level or at the altitude of the -10 C isotherm) is never met. Yet, convective clouds do in fact grow to produce rainfall. This leads to the notion that "you know it when you see it" in terms of CI, but in these situations a hard radar-based definition is less easy to follow.



In this example, the SATCAST (proxy Official AWG CI Algorithm) flagged a cluster of cumulus clouds at 1845 UTC, which are seen in 1 km GOES-13 visible satellite data at 1932 UTC. The first radar echo was seen near 1924 UTC with a maximum dBZ of <20 dBZ. This small cell developed over the Plains, and propagated north-northeatward (to the northeast of Trinidad, CO), and maintained only 10-20 dBZ rainfall. Radar data are shown through 2000 UTC. Also shown is the RUC sounding for near this location which confirm (a) the storm occurred in a dry easterly flow, to the north of a warm from, (b) most unstable CAPE values were near 471 Jkg^-1, and (c) cloud bases were near 680 hPa, with storm tops perhaps reaching ~300 hPa (i.e. it was elevated convection).

This example highlights the sensitivity of the SATCAST algorithm for identifying CI in less than obvious situations (when CAPE values are not high, and when cumulus updraft widths do not necessarily fill a 4 km^2 GOES pixel).

pGLM realtime comparison with MESH and 3D-Var Updraft

Repost from EWP blog...

With warning operations already underway for western Oklahoma, forecasters are deep into their storm analysis.

One of the more interesting features they have been picking up on is the consistent signals between the pGLM lightning trends and values from the MESH (Maximum Expected Size of Hail) algorithm as well as the 3D-Var derived updraft fields.

This was well illustrated by the storm north of Elk City moving from Beckham to Roger Mills county. At approximately the same time ~1900-1915 UTC, the lightning rate increased from 5 to 15 flashes per min (per pGLM grid box, not per storm) as MESH ramped up and updraft increased within the 3D-Var product. Shortly after this increase, both the pGLM and MESH values decreased with this storm (the 3D-Var updraft values also showed this, but with a bit of a time lag).
pGLM flash density and MESH values at 1910 UTC on 19 May 2011
Also of note, prior to losing NLDN data, with the storms seemed to be producing relatively little CG lightning. In this case, the pGLM data was definitely giving a better view of the electrical activity and storm intensity. (1 to 1.75 in hail has already been reported across West and SW Oklahoma).

-K. Kuhlman (pGLM scientist, week 2)