Compare Threatnet Accuracy

[Bill Tabor]

A lot of us often complain about the lack of timeliness and accuracy of Mobile Threatnet. I've had a few occasions where I saw a core on the screen directly ahead only to have it visually on the right side of the road already. My question though is how does it compare to Grlevelx, or Weathertap and similar services we are all used to having. I've been assuming Threatnet isn't as accurate as Grlevelx placement wise but we may be surprised.

Today I did a short chase about 20 miles round trip and when I pulled up in the driveway and obtained wifi with Threatnet still running I was in the middle of a hailstorm of nickel hail. I stayed in the vehicle cause it looked like those things would hurt. Eventually pulled partway in to the garage and after a minute or so compared Threatnet and Grlevelx. I believe it had been a few minutes at least when I looked and sure enough Grlevelx showed the storm to the east and Threatnet showed it still overhead. I then noticed the activity light was red. Oh yeah, you have to have a line of sight to the sky for Threatnet. So I backed up in the driveway and after a few minutes it had updated. I compared the two current images while zoomed in showing the towns of Cedar Park and Leander as well as the Williamson county line and Hwy 183. Both showed my vehicle just to the west of the storm echo which had just left. Seemed that Grlevelx had a better more formed shaped as opposed to a rounder more 'blobbier' shape of Threatnet, but they seemed to be lined up mostly over the same real estate.

So, I believe based on that one test they may be similar. What are your results? I'd like to have a number of us test this preferably with GPS, but running Threatnet and another leading package while looking at a storm and see if it is accurate, how much time shifted, and if the two radars show the precip in similar locations. I believe any timing errors should be magnified with faster moving storm motions.

 

[Jason A.C. Brock]

A lot of us often complain about the lack of timeliness and accuracy of Mobile Threatnet. I've had a few occasions where I saw a core on the screen directly ahead only to have it visually on the right side of the road already. My question though is how does it compare to Grlevelx, or Weathertap and similar services we are all used to having. I've been assuming Threatnet isn't as accurate as Grlevelx placement wise but we may be surprised.

Today I did a short chase about 20 miles round trip and when I pulled up in the driveway and obtained wifi with Threatnet still running I was in the middle of a hailstorm of nickel hail. I stayed in the vehicle cause it looked like those things would hurt. Eventually pulled partway in to the garage and after a minute or so compared Threatnet and Grlevelx. I believe it had been a few minutes at least when I looked and sure enough Grlevelx showed the storm to the east and Threatnet showed it still overhead. I then noticed the activity light was red. Oh yeah, you have to have a line of sight to the sky for Threatnet. So I backed up in the driveway and after a few minutes it had updated. I compared the two current images while zoomed in showing the towns of Cedar Park and Leander as well as the Williamson county line and Hwy 183. Both showed my vehicle just to the west of the storm echo which had just left. Seemed that Grlevelx had a better more formed shaped as opposed to a rounder more 'blobbier' shape of Threatnet, but they seemed to be lined up mostly over the same real estate.

So, I believe based on that one test they may be similar. What are your results? I'd like to have a number of us test this preferably with GPS, but running Threatnet and another leading package while looking at a storm and see if it is accurate, how much time shifted, and if the two radars show the precip in similar locations. I believe any timing errors should be magnified with faster moving storm motions.
[/b]

I have had no problems as far as threatnet showing me exactly where the cells are. You have to be careful about pulling under overpasses or gas station awnings etc. because it blocks the XM system Xm radios are jsut like this and are basically the same signal.
The only thing I dislike about threatnet is basically it is not nearly as good a resoluton of the storm as grlevel3 etc. Threatnet will show shear markers and areas of shear but they do not show features well enough to make out hook echos or gustfronts etc. The shear areas areas semm to be more conducive to show areas of large hail actually. The shear markers showing rotation due give a nice idea of where rotation can be and usaully does on very tornadic type cells but you can also get these shear markers pop up in areas where there may be more of a high wind threat than a tornadic threat. Especially on the leading edges of intense squall lines and or bowing type storms.
I have evn had areas of shear show up in areas of relatively weak to moderate rain and the system is probably picking up outflow or even heat bursts.
Anyways if grlevel 3 was available like threat net and had nationwide coverage and updated at least every 5 minutes was able to be received at ALL times I would jump on that in a second. WX WORX really really needs to make its radar more on the lines of the resolution grlevel 3 is. they also nned to do this in a hurry. its only a matter of time when we will have sattelite internet that works EVERYWHERE and at ALL TIMES. In fact their is a product I know of that will be available in late summer but the price for now is very high. give it a couple years folks.......its coming. Total and complete high speed mobile internet. Hopefully the cellular providers wont latch on to it and bleed us for outrages prices.

Also if you have problems with the threatnet signal out in the open you can go to your XM account at xmradio.com and refresh the signal or call them at 1800xmradio. XM radios and the marine type weather setups take about 15-20 minutes to get the refresh signal which is sent out for 36 hours after you ask them to do it. The responder package with all the products however can take as long as 3 hours to get the refresh signal since it is receiving so much data products. If you plan to chase its best to test your system the night before at least or even in the morning so that if there happens to be a problem you can get it resolved before the chase. if ya have any other issues the customer supposrt at Baron Wxworx are very helpful and can talk you through almost anything. If you cave COM errors its usually becuase of a loose cable. Especially the USB however now they have bluetooth but I didnt want to upgrade. A few extra wires arent the inconvenience of a 200 dollar upgrade.

 

[David Drummond]

I may be wrong, someone that knows better please correct this, but isn't WXWORX composite radar data from all sites and all tilts? That being the case, even with more resolution does that not prevent you from usually determining low level scan features such as hook echos, BWER etc etc? Add in the smoothing and definitely so.

Also keep in mind, you aren't getting that radar data live, with WXWORX or GRLevelX. By the time the radar finishes it's scan, (at the fastest VCP 5 minutes) and the data is processed and you get it, your may be looking at 8 to 10 minutes old. Given a storm moving at 30 mph, it could have moved a good 5 miles over from the time the scan was taken until you get the data.

This might be faster with data from places like AllisonHouse.com....maybe Tyler could comment on how "old" the data is there, but I am definitely sure it takes a bit longer with WXWORX with the satellites involved etc.

 

[Shane Adams]

Mickey Ptak and I noted several instances last Thursday where threat net/wx-worx or whatever was inaccurate. We'd have cores east of us and the display still had the core over our heads. Also, twice it went down completely, at critical points in the chase when we were actually using it to judge approach vectors when time was of the essence. Wx-worx is a fun tool and it gives you something to do waiting for initiation, but in the end it's merely a guide (like nowcasters). IMO it's much more valuable to inexprienced chasers who are more apt to stare at the screen more than the sky; putting blind faith in wx-worx guidence can lead a chaser to the right spot if it's accurate data. But for seasoned chasers who are more likely to look at the sky, it's just a fun sidekick that provides some entertainment along the way. The trick with this or any guidence software is to use it conservatively. Switching off the radio or scanner or your own eyes can be a costly mistake, and even if you see a tornado, what have you learned? You've learned that guidence software is AWESOME!!! And so the cycle continues.

Just opinions.

 

[Tyler Allison]

This might be faster with data from places like AllisonHouse.com....maybe Tyler could comment on how "old" the data is there, but I am definitely sure it takes a bit longer with WXWORX with the satellites involved etc.
[/b]

I don't have a WxWorx so I can't do any tests but you can't get it any faster than AllisonHouse unless you have a T1 into the NWS. Baron/etc.. have them. I'm working on it.

Even with a T1 into the NWS, Baron (and others) have to run the data through processing engines, package up the data and ship it out to a satellite uplink. That has to take at least a few seconds...maybe up to 60..but I have no idea.

However, I doubt there would be anything significant if you compared the two. I could explain why I think that but that's not the point of this thread.

 

[David Drummond]

That's about what I figured, so realistically then Tyler, with your feed, what we see is probably less than 7 minutes old from the time that particular radar scan started (in highest VCP mode)?


I personally use Tyler's feed with GRL3 and just leave the data feed part of WXWORX open and close the map, and then when I am out of my regular data coverage to get his feeds, I switch over to WXWORX.

Like Shane says though, it's only a tool. It's best use is "pre-game" up until you have locked on your target storm and are in position. Once your there 95% of the time the storm will visually tell you what you need to know although looking at a "live" radar and looking at the storm can seriously help you visualize what a storm looks like when your stuck at home watching an outbreak 7 states away. :blink:

 

[Jeff Snyder]

David,

A full volume scan in VCP12 takes like 4:20 to complete. However, if I'm not mistaken, each tilt is transmitted seperately, a few minutes after that elevation angle scan is complete. So, it's not like we don't see tilt 1 until all tilts are completed, processed, and transmitted. Again, AFAIK, once the lowest elevation angle is complete, the data are processed and transmitted, all while the next several elevation angle scans are being collected. I would think that the data are sent out (and we receive them) 2-3 minutes after the scan is complete. On VCPs 11 and 12, antenna rotation rate is rapid, so the actual time spent during a scan at a particular elevation angle is pretty short (remember also that the lowest few elevation angles are actually scanned twice, once with a low PRF for longer range reflectivity, followed by a scan at the same elevation level with a high PRF, used to gather velocity data / boost the aliasing/Nyquist velocity).

 

[David Drummond]

Jeff, do you know when the data is time stamped? Is it when the scan is complete and the data is sent? Or when the scan is started? Or when the end software assembles the data into an end product?

 

[Jeff Snyder]

Jeff, do you know when the data is time stamped? Is it when the scan is complete and the data is sent? Or when the scan is started? Or when the end software assembles the data into an end product?
[/b]

David,

On GR3 at least, there are two times given -- VST (Volume scan Start Time, though GR3 picks the time that the elevation scan started, IIRC) and Prod (product generation time). I guess the two times (which are based on two different clocks) are not always accurate, since the two clocks may not be synched. Regardless, I almost always seem to get the data 2-4 minutes after the Prod time (and 3-5 mins after the VST time). It seems that I can get images ~30-45s earlier on some internet sites (like CoD), but that's probably the difference between slower NWS FTP servers (which is noticeable on big severe weather days) and faster gateways (like NOAAPORT / direct T1 link / etc). I'm sure there are others who know more abt the time stamps on radar data, so maybe they'll respond. And of course, that's the fastest I can get the data... If I look at the image when it's 3 minutes old, the actual lag may be 3min (time from the scan to my computer) + 3min (how long I've had it up on screen), or 6 minutes old. For storms that are moving 50-60mph, such a delay may be significant.


EDIT: As noted below, the composite reflectivity product is one of the last to be made, since all elevation scans must be completed first. For VCP12, this means that the comp refl image may contain data that are ~4 minutes old for the get-go, before generation/processing and transmission is taken into consideration.

 

[GPhillips]

Looking at GRLevel3 data at home, the 0.5 degree elevation slices typically come in about 2 minutes after the volume scan start time, using the NWS servers as the data source. Mike lists both the start of the volume scan (VST), and the start of the elevation angle (Prod) in the GRLevel3 legend.

Time stamps for products elsewhere are generally given as the start of the volume scan on images at such places as NWS home pages and college of DuPage radar.

If the reflectivity type product used in threatnet is a composite image for a volume scan that matches the algorithm derived shear product, the earliest this product would come in would be 5 to 7 minutes after the start of the volume scan when the radars are in precipitation mode.

 

[GPhillips]

On GR3 at least, there are two times given -- VST (Volume scan Start Time, though GR3 picks the time that the elevation scan started, IIRC) and Prod (product generation time). I guess the two times (which are based on two different clocks) are not always accurate, since the two clocks may not be synched. [/b]

This is possible until the OpenRDA is installed. The Product times should have been accurate to within a few seconds for the last year or so since the Radar Product Generator has been time synced to AWIPS. The VST or Volume Scan Start Time was read from the RDA. The RDA clock was not automatically time synched (given the old hardware), but manually checked and reset once a month. With the OpenRDA which is going in across the country now...Installation map here, the RDA time, and thus the VST time, is synched via GPS. Looking at the radars where the OpenRDA has been installed, the VST and Prod times on GRLevel3 are no more than 3 seconds apart, and the Prod time is always later than the VST. So these potential time discrepancy problems should be going away.

 

[Greg Stumpf]

...each tilt is transmitted seperately, a few minutes after that elevation angle scan is complete.[/b]

Live LevelII data is transmitted on the CRAFT network in 100 radial packets, so conceivably, you can process partial elevation angle before they're in.

Based on partial information I have of the MTN reflectivity, it is a 2 km grid of maximum reflectivity in the vertical (aka "composite reflectivity"), mosaicked from the latest data from multiple radars every 5 minutes. Not sure if they are using "virtual volume scans", but if not, data from one particular radar could conceivably be up to 11 minutes old (and then add download and post-processing time), if the 5 minute update comes right at the end of a 6 minute volume scan for one of the radars. The gridded data are also dumbed down to 7 reflectivity levels, and then recombined and OBANed after data tranmission from XM into the cartoonish grid we see on the display.

BTW - Jeff Piotrowski wouldn't be specific at the Denver Convention, but said there were some improvements in the offing this spring. I've so far heard of none so far this year. Does anyone have any more info? I do see some "new" stuff in their Threat Net Plus, but I believe this is not the mobile package.

 

[Dan Robinson]

When chasing with WxWorx you have to learn to adjust to the fact that the radar is 10 minutes behind, and plan accordingly. Particularly with fast-moving storms. It is tricky but can be done. For instance Friday, when we saw a core on radar to the left of the road, we knew it was probably already well across on the right side.

The shear SCITS and 'hail' tracker points update much more rapidly and are more current, so it is possible to rely on those for positioning more so than the radar image, especially with an established supercell.